Landscapes of (Co)Evolution: A Tale of Two Signals
by
Victoria R. Caudill
A dissertation accepted and approved in partial fulfillment of the 
requirements for the degree of
Doctor of Philosophy
in Biology
Dissertation Committee:
Dr. Brendan Bohannan, Chair     
Peter Ralph, Advisor
Dr. Matt Barber, Core Member
Dr. Lauren Ponisio, Core Member
Dr. Mike Harms, Institutional Representative
University of Oregon
Spring 2024
© 2024 Victoria R. Caudill
2
DISSERTATION ABSTRACT
Victoria R. Caudill
Doctor of Philosophy in Biology
Title: Landscapes of (Co)Evolution: A Tale of Two Signals
Evolution between species and within specific environments has resulted in a
diverse array of traits and genetic variation. These are the result of interactions
that have occurred across space and throughout time. As evolution progresses, it
generates signals and patterns that can be used to unravel mysteries of the past
and provide insight into future possibilities. By examining the signals left behind
by the process of evolution, I have gained valuable insights onto what has occurred
in the past. In chapter II, I use spatial simulations to explore the (co)evolutionary
trajectories of levels of toxin resistance and toxin production in the predator-prey
Thamnophis garter snake – Taricha newt system. Specifically, I examine how possi-
ble genetic architectures of the toxin and resistance traits affect the coevolutionary
dynamics by manipulating both mutation rate and effect size of mutations across
many simulations. I find that coevolutionary dynamics alone were not sufficient in
our simulations to produce the striking mosaic of levels of toxicity and resistance
observed in nature. Instead, simulations with ecological heterogeneity (in trait cost-
liness or interaction rate) did produce such patterns. In chapter III, I examined
landscapes of genetic variation in cichlids, a species complex that has recently ra-
diated. I used the phylogenetic relationship and population genetic measurements
(mean nucleotide diversity and divergence) to describe large-scale variation across
the genome. These patterns are likely caused by complex effects of inversions, in-
trogression, and linked selection. Together, these findings contribute to building a
strong foundation for understanding the evolutionary signals of natural selection
and how its’ impacts vary along the genome. This dissertation includes previously
published and unpublished co-authored material.
3
TABLE OF CONTENTS
Chapter Page
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2. CHAPTER 2 GENETIC ARCHITECTURE, SPATIAL HETEROGENEITY,
AND THE COEVOLUTIONARY ARMS RACE BETWEEN NEWTS AND
SNAKES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
The demographic model . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Genetic Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Simulation Experiments . . . . . . . . . . . . . . . . . . . . . . . . . 21
The contribution of coevolution to phenotype change . . . . . . . . . . 22
Heterogeneous landscapes . . . . . . . . . . . . . . . . . . . . . . . . . 24
Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Data availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Newt and snake Evolution . . . . . . . . . . . . . . . . . . . . . . . . 25
How much is coevolution driving phenotype change? . . . . . . . . . . 27
Spatially Heterogeneous Landscapes . . . . . . . . . . . . . . . . . . . 28
Effects of Genetic Architecture on Coevolution . . . . . . . . . . . . . 32
2.4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Relationship to Other Models of Coevolution . . . . . . . . . . . . . . 36
What Does This Tell us About the Newts and Snakes? . . . . . . . . . 37
Spatial Patterns and Interactions . . . . . . . . . . . . . . . . . . . . 38
Genetic Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Limitations and Continuing Questions . . . . . . . . . . . . . . . . . 41
2.5. Bridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4
3. CHAPTER 3 THE GENETIC LANDSCAPES OF LAKE MALAWI CICH-
LIDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Samples and Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Phylogeny . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Population genetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Introgression Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Divergence across the genome . . . . . . . . . . . . . . . . . . . . . . 56
Estimating phylogenetic distance . . . . . . . . . . . . . . . . . . . . . 58
Correlations between landscapes . . . . . . . . . . . . . . . . . . . . . 58
Correlations with genomic features . . . . . . . . . . . . . . . . . . . 61
Introgression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Features of genomic landscapes . . . . . . . . . . . . . . . . . . . . . 66
Chromosomal rearrangements . . . . . . . . . . . . . . . . . . . . . . 66
Sex determining regions . . . . . . . . . . . . . . . . . . . . . . . . . 67
Hybridization and introgression . . . . . . . . . . . . . . . . . . . . . 68
Limitations and continuing questions . . . . . . . . . . . . . . . . . . 68
3.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
APPENDICES
A. COEVOLUTION APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . 78
B. CICHLID APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
C. WINDOW DXY BY SPECIES . . . . . . . . . . . . . . . . . . . . . . . . 94
D. WINDOW π BY SPECIES . . . . . . . . . . . . . . . . . . . . . . . . . . 109
E. CM BY BASE PAIR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
F. CORRELATION AND COVARIANCE BY LINKAGE GROUPS . . . . . 136
5
G. CORRELATION OF dXY AND GENOME FUNCTIONS FOR ALL LINK-
AGE GROUPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
H. PHYLOGENETIC TREES BY LINKAGE GROUP . . . . . . . . . . . . . 148
I. FBRANCH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6
LIST OF FIGURES
Figure Page
2.1. Mean phenotype dynamics over time in simulations showing three speeds
of evolution; fast, slow, and no. . . . . . . . . . . . . . . . . . . . . . . . 26
2.2. Mean phenotypes for each combination of genetic architectures in Exper-
iment 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3. Newt and snake mean phenotypes, comparing standard simulations to sim-
ulations without heritability or random interaction outcomes . . . . . . . 29
2.4. Spatial phenotype correlations for three types of map . . . . . . . . . . . 30
2.5. Local newt and snake phenotypes across the geographical area in a simu-
lation in which costliness of the phenotype was a spatial gradient . . . . 31
2.6. Speed of coevolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.7. Distribution of newt and snake phenotypes and population sizes . . . . . 35
3.1. An example phylogenetic tree . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2. Divergence in 1Mb windows between all species pairs on all linkage groups 57
3.3. Correlations between different landscapes of genetic diversity or divergence
plotted against separating phylogenetic time . . . . . . . . . . . . . . . . 59
3.4. Shows the correlation between four genome functions (gene density, gene re-
peats, accessibility, recombination) and dXY for each linkage group . . . 62
3.5. Average fd for 1Mb windows across the entire genome . . . . . . . . . . . 64
A.1. Early population size and phenotype for newts and snakes . . . . . . . . 79
A.2. Late population size and phenotype for newts and snakes . . . . . . . . . 80
A.3. Spatial phenotype correlation for additional backgrounds . . . . . . . . . 81
A.4. Phenotype Differences by Mutational Variance for experiment 2 . . . . . 82
A.5. Phenotype Differences by Mutational Variance for experiment 3 . . . . . 83
B.1. Phylogenetic tree of all individuals . . . . . . . . . . . . . . . . . . . . . 88
B.2. Genetic diversity in 1Mb windows . . . . . . . . . . . . . . . . . . . . . . 89
B.3. Whole-genome phylogenetic tree for one individual per species . . . . . . 90
B.4. Whole genome correlations for all species pairs. . . . . . . . . . . . . . . 91
B.5. Correlations of landscapes of diversity and divergence in the great apes with
recombination rate and exon density . . . . . . . . . . . . . . . . . . . . 92
7
B.6. 1Mb window of admixture proportions . . . . . . . . . . . . . . . . . . . 93
C.1. 1Mb window dXY for all linkage groups, C.intermedius . . . . . . . . . . 95
C.2. 1Mb window dXY for all linkage groups, M.anaphyrmus . . . . . . . . . . 96
C.3. 1Mb window dXY for all linkage groups, P.subocularis . . . . . . . . . . . 97
C.4. 1Mb window dXY for all linkage groups, C.rhoadesii . . . . . . . . . . . . 98
C.5. 1Mb window dXY for all linkage groups, C.caeruelus . . . . . . . . . . . . 99
C.6. 1Mb window dXY for all linkage groups, D.strigatus . . . . . . . . . . . . 100
C.7. 1Mb window dXY for all linkage groups, P.ornatus . . . . . . . . . . . . . 101
C.8. 1Mb window dXY for all linkage groups, F.rostratus . . . . . . . . . . . . 102
C.9. 1Mb window dXY for all linkage groups, H.oxyrhynchus . . . . . . . . . . 103
C.10.1Mb window dXY for all linkage groups, L.lethrinus . . . . . . . . . . . . 104
C.11.1Mb window dXY for all linkage groups, C.virginalis . . . . . . . . . . . . 105
C.12.1Mb window dXY for all linkage groups, O.speciosus . . . . . . . . . . . . 106
C.13.1Mb window dXY for all linkage groups, T.placodon . . . . . . . . . . . . 107
C.14.1Mb window dXY for all linkage groups, P.longimanus . . . . . . . . . . . 108
D.1. 1Mb window π for all linkage groups, C.intermedius . . . . . . . . . . . . 110
D.2. 1Mb window π for all linkage groups, M.anaphyrmus . . . . . . . . . . . 111
D.3. 1Mb window π for all linkage groups, P.subocularis . . . . . . . . . . . . 112
D.4. 1Mb window π for all linkage groups, C.rhoadesii . . . . . . . . . . . . . 113
D.5. 1Mb window π for all linkage groups, C.caeruelus . . . . . . . . . . . . . 114
D.6. 1Mb window π for all linkage groups, D.strigatus . . . . . . . . . . . . . 115
D.7. 1Mb window π for all linkage groups, P.ornatus . . . . . . . . . . . . . . 116
D.8. 1Mb window π for all linkage groups, F.rostratus . . . . . . . . . . . . . 117
D.9. 1Mb window π for all linkage groups, H.oxyrhynchus . . . . . . . . . . . 118
D.10.1Mb window π for all linkage groups, L.lethrinus . . . . . . . . . . . . . 119
D.11.1Mb window π for all linkage groups, C.virginalis . . . . . . . . . . . . . 120
D.12.1Mb window π for all linkage groups, O.speciosus . . . . . . . . . . . . . 121
D.13.1Mb window π for all linkage groups, T.placodon . . . . . . . . . . . . . 122
D.14.1Mb window π for all linkage groups, P.longimanus . . . . . . . . . . . . 123
E.1. cM by base pair linkage group one . . . . . . . . . . . . . . . . . . . . . 124
E.2. cM by base pair linkage group two . . . . . . . . . . . . . . . . . . . . . 125
E.3. cM by base pair linkage group three . . . . . . . . . . . . . . . . . . . . . 125
E.4. cM by base pair linkage group four . . . . . . . . . . . . . . . . . . . . . 126
8
E.5. cM by base pair linkage group five . . . . . . . . . . . . . . . . . . . . . 126
E.6. cM by base pair linkage group six . . . . . . . . . . . . . . . . . . . . . . 127
E.7. cM by base pair linkage group seven . . . . . . . . . . . . . . . . . . . . 127
E.8. cM by base pair linkage group eight . . . . . . . . . . . . . . . . . . . . . 128
E.9. cM by base pair linkage group nine . . . . . . . . . . . . . . . . . . . . . 128
E.10.cM by base pair linkage group ten . . . . . . . . . . . . . . . . . . . . . . 129
E.11.cM by base pair linkage group eleven . . . . . . . . . . . . . . . . . . . . 129
E.12.cM by base pair linkage group twelve . . . . . . . . . . . . . . . . . . . . 130
E.13.cM by base pair linkage group thirteen . . . . . . . . . . . . . . . . . . . 130
E.14.cM by base pair linkage group fourteen . . . . . . . . . . . . . . . . . . . 131
E.15.cM by base pair linkage group fifteen . . . . . . . . . . . . . . . . . . . . 131
E.16.cM by base pair linkage group sixteen . . . . . . . . . . . . . . . . . . . . 132
E.17.cM by base pair linkage group seventeen . . . . . . . . . . . . . . . . . . 132
E.18.cM by base pair linkage group eighteen . . . . . . . . . . . . . . . . . . . 133
E.19.cM by base pair linkage group nineteen . . . . . . . . . . . . . . . . . . . 133
E.20.cM by base pair linkage group twenty . . . . . . . . . . . . . . . . . . . . 134
E.21.cM by base pair linkage group twenty-two . . . . . . . . . . . . . . . . . 134
E.22.cM by base pair linkage group twenty-three . . . . . . . . . . . . . . . . 135
F.1. Correlation between π and π in 1Mb widowed by phylogenetic time, all link-
age groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
F.2. Covariance between π and π in 1Mb widowed by phylogenetic time, all link-
age groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
F.3. Correlation between dXY and π in 1Mb widowed by phylogenetic time, all
linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
F.4. Covariance between dXY and π in 1Mb widowed by phylogenetic time, all
linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
F.5. Correlation between dXY and dXY in 1Mb widowed by phylogenetic time,
all linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
F.6. Covariance between dXY and dXY in 1Mb widowed by phylogenetic time,
all linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
G.1. Correlation between dXY and gene repeats in 1Mb widowed by phylogenetic
time, all linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9
G.2. Correlation between dXY and gene density in 1Mb widowed by phylogenetic
time, all linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
G.3. Correlation between dXY and accessibility in 1Mb widowed by phylogenetic
time, all linkage groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
G.4. Correlation between dXY and recombination rate in 1Mb widowed by phy-
logenetic time, all linkage groups . . . . . . . . . . . . . . . . . . . . . . 147
H.1. Phylogenetic tree, linkage group 1 . . . . . . . . . . . . . . . . . . . . . . 149
H.2. Phylogenetic tree, linkage group 2 . . . . . . . . . . . . . . . . . . . . . . 150
H.3. Phylogenetic tree, linkage group 3 . . . . . . . . . . . . . . . . . . . . . . 151
H.4. Phylogenetic tree, linkage group 4 . . . . . . . . . . . . . . . . . . . . . . 152
H.5. Phylogenetic tree, linkage group 5 . . . . . . . . . . . . . . . . . . . . . . 153
H.6. Phylogenetic tree, linkage group 6 . . . . . . . . . . . . . . . . . . . . . . 154
H.7. Phylogenetic tree, linkage group 7 . . . . . . . . . . . . . . . . . . . . . . 155
H.8. Phylogenetic tree, linkage group 8 . . . . . . . . . . . . . . . . . . . . . . 156
H.9. Phylogenetic tree, linkage group 9 . . . . . . . . . . . . . . . . . . . . . . 157
H.10.Phylogenetic tree, linkage group10 . . . . . . . . . . . . . . . . . . . . . 158
H.11.Phylogenetic tree, linkage group 11 . . . . . . . . . . . . . . . . . . . . . 159
H.12.Phylogenetic tree, linkage group 12 . . . . . . . . . . . . . . . . . . . . . 160
H.13.Phylogenetic tree, linkage group 13 . . . . . . . . . . . . . . . . . . . . . 161
H.14.Phylogenetic tree, linkage group 14 . . . . . . . . . . . . . . . . . . . . . 162
H.15.Phylogenetic tree, linkage group 15 . . . . . . . . . . . . . . . . . . . . . 163
H.16.Phylogenetic tree, linkage group 16 . . . . . . . . . . . . . . . . . . . . . 164
H.17.Phylogenetic tree, linkage group 17 . . . . . . . . . . . . . . . . . . . . . 165
H.18.Phylogenetic tree, linkage group 17 . . . . . . . . . . . . . . . . . . . . . 166
H.19.Phylogenetic tree, linkage group 18 . . . . . . . . . . . . . . . . . . . . . 167
H.20.Phylogenetic tree, linkage group 19 . . . . . . . . . . . . . . . . . . . . . 168
H.21.Phylogenetic tree, linkage group 20 . . . . . . . . . . . . . . . . . . . . . 169
H.22.Phylogenetic tree, linkage group 22 . . . . . . . . . . . . . . . . . . . . . 170
H.23.Phylogenetic tree, linkage group 23 . . . . . . . . . . . . . . . . . . . . . 171
I.1. fbranch result whole genome . . . . . . . . . . . . . . . . . . . . . . . . . 173
I.2. fbranch result linkage group 1 . . . . . . . . . . . . . . . . . . . . . . . . 174
I.3. fbranch result linkage group 2 . . . . . . . . . . . . . . . . . . . . . . . . 175
I.4. fbranch result linkage group 3 . . . . . . . . . . . . . . . . . . . . . . . . 176
10
I.5. fbranch result linkage group 4 . . . . . . . . . . . . . . . . . . . . . . . . 177
I.6. fbranch result linkage group 5 . . . . . . . . . . . . . . . . . . . . . . . . 178
I.7. fbranch result linkage group 6 . . . . . . . . . . . . . . . . . . . . . . . . 179
I.8. fbranch result linkage group 7 . . . . . . . . . . . . . . . . . . . . . . . . 180
I.9. fbranch result linkage group 8 . . . . . . . . . . . . . . . . . . . . . . . . 181
I.10. fbranch result linkage group 9 . . . . . . . . . . . . . . . . . . . . . . . . 182
I.11. fbranch result linkage group 10 . . . . . . . . . . . . . . . . . . . . . . . 183
I.12. fbranch result linkage group 11 . . . . . . . . . . . . . . . . . . . . . . . 184
I.13. fbranch result linkage group 12 . . . . . . . . . . . . . . . . . . . . . . . 185
I.14. fbranch result linkage group 13 . . . . . . . . . . . . . . . . . . . . . . . 186
I.15. fbranch result linkage group 14 . . . . . . . . . . . . . . . . . . . . . . . 187
I.16. fbranch result linkage group 15 . . . . . . . . . . . . . . . . . . . . . . . 188
I.17. fbranch result linkage group 16 . . . . . . . . . . . . . . . . . . . . . . . 189
I.18. fbranch result linkage group 17 . . . . . . . . . . . . . . . . . . . . . . . 190
I.19. fbranch result linkage group 18 . . . . . . . . . . . . . . . . . . . . . . . 191
I.20. fbranch result linkage group 19 . . . . . . . . . . . . . . . . . . . . . . . 192
I.21. fbranch result linkage group 20 . . . . . . . . . . . . . . . . . . . . . . . 193
I.22. fbranch result linkage group 22 . . . . . . . . . . . . . . . . . . . . . . . 194
I.23. fbranch result linkage group 23 . . . . . . . . . . . . . . . . . . . . . . . 195
11
LIST OF TABLES
Table Page
2.1. A summary of all parameter sets used in simulations. . . . . . . . . . . . 23
B.1. Samples used in this study (from Malinsky, Svardal, et al. (2018)) . . . . 85
B.1. The table (continued from previous page) . . . . . . . . . . . . . . . . . . 86
B.1. The table (continued from previous page) . . . . . . . . . . . . . . . . . . 87
12
CHAPTER 1
INTRODUCTION
The study of evolution is complex and messy, leaving multiple threads for re-
searchers to grasp. At its core, evolution is the unifying principle of biology that
explains how life on Earth has diversified and adapted over billions of years. Evolu-
tion is mediated by several different mechanisms including natural selection, adap-
tation, mutation, gene flow, genetic drift, non-random mating, and genetic varia-
tion. These mechanisms can interact in various ways, leading to diverse evolution-
ary outcomes. Furthermore, evolution can act over multiple timescales—from short-
term changes within populations (microevolution) to long-term patterns shaping
the diversity of life over geological epochs (macroevolution). The process of evolu-
tion is also intricately linked with environmental factors, such as climate change,
habitat alteration, and species interactions, which shape the selective pressures act-
ing on organisms.
The foundation of the study of evolution began with simple observations about
the relationships between organisms. Over time, these observations have grown into
theories and experiments that dissect the intricate relationships between these dif-
ferent mechanisms and timescales of evolution. Gradually, our understanding of
evolution has grown, as the field of genetics has been integrated with evolution-
ary theory. From the many observations conducted, experiments ran, and theory
produced, evolutionary biology has become an umbrella with various areas of fo-
cus that are encompassed by this term (i.e., speciation, developmental biology,
phylogenetics, coevolution, systematics, population genetics, etc..). Each of these
subfields have several ways of describing and quantifying signals of evolutionary
change. Through there are various fields in evolutionary biology, in this disserta-
tion I will focus on coevolution and population genetics. Researchers have explored
the reciprocal evolutionary changes between species interacting within their envi-
ronment and with eachother. These interactions, also referred to as coevolution,
can take place in wide networks consisting of multiple species interacting in various
environments across space and time. Understanding these complex interactions is
fundamental in providing insights into the shaping of ecological communities and
their impacts on biodiversity. Studies on coevolution have mainly focused on host-
parasite relationships, plant-pollinator associations, and predator-prey interactions.
In chapter II, I study evolutionary trajectories in a coevolutionary arms race of a
13
predator-prey (snake-newt) system. Using simulation, I explore the geographic vari-
ation of newts’ toxicity and snakes’ resistance as these species interact across space
and time. This study focuses on the effects of genetic basis on trait evolution, in
the context of realistic reciprocal selection, genetic drift, gene flow, and spatial en-
vironmental variation. This work is currently in review and is co-authored with Dr.
Peter Ralph.
Evolutionary biology also encompasses the study of common ancestry, where
all existing organisms are related through decent from a common ancestor over the
course of evolutionary history. As species evolve, different evolutionary mechanisms
leave lasting signals within their genomes. These signals are often difficult to dis-
cern due to evolutionary mechanisms occurring at different times, in opposition,
or jointly. Oftentimes, evolution impacts different areas of a genome differently. In
chapter III, I examine the signals left by evolution along the genome of 14 species
of cichlid. We found that there are similarities in the peaks and valleys of the land-
scapes of genetic diversity and divergence of these species. These genetic landscapes
are highly correlated for sister species but are less correlated for more phylogenet-
ically distant species. There are also major differences between the linkage groups,
suggesting multiple evolutionary mechanisms acting differently along the genome.
This work is unpublished and is co-authored with Dr. Anastasia Teterina, Aidan
Short, and Dr. Peter Ralph.
Extensive studies of evolution have greatly advanced our understanding of
adaptation, genetics, and diversity. My work in this dissertation continues this ex-
pansion of our knowledge by investigating some of the many signals that are pro-
duced by evolutionary processes.
14
CHAPTER 2
CHAPTER 2 GENETIC ARCHITECTURE, SPATIAL HETEROGENEITY, AND
THE COEVOLUTIONARY ARMS RACE BETWEEN NEWTS AND SNAKES
This work is currently in review. The experimental question and design for
this project was developed by myself and Dr. Peter L. Ralph. I ran and analyzed
all simulation results and wrote up the manuscript for this project jointly with Dr.
Peter L. Ralph.
2.1 Introduction
Coevolution is the reciprocal adaptation of heritable traits between interact-
ing species (Janzen et al. 1980). These dynamic interactions between species shape
patterns of adaptation, genetic diversity, and ecological dynamics. Coevolution
shapes our world through mutualistic partnerships (such as pollinators and plants)
(Janzen 1966), symbiotic dependencies (Thrall et al. 2007) and arms races (e.g.,
between host and pathogen or predator and prey) (Daugherty and Malik 2012).
These multifarious interactions not only contribute to biodiversity, but also define
species relationships, and accentuate the development of exaggerated traits. A bet-
ter understanding of coevolution will help us to better craft conservation strategies,
efficiently manage pest populations, and understand the dynamics that underpin
stable ecosystems.
A coevolutionary interaction between species in which the individuals of one
species benefit from the interaction and the individuals of the other is being harmed
or killed and where phenotypes are continuously escalating is called an antagonistic
coevolutionary arms race. For instance, a prey might evolve a defense mechanism
and in response, its predator might respond by evolving a counter-mechanism, re-
sulting in an ongoing cycle of adaptations and counter-adaptations (e.g., between
herbivore resistance and plant defenses, Ehrlich and Raven 1964). Arms races be-
tween species have been observed in many species pairs such as Taricha newts
and Thamnophis garter snakes (Brodie 2003), bacteria and phage (Bohannan and
Lenski 2000), flax and flax rust (Dodds et al. 2006), and parsnip and parsnip web-
worm (Berenbaum, Zangerl, and Nitao 1986).
Antagonistic coevolution can be complex, especially when multiple interac-
tions occur across space and time (Forde, Thompson, and Bohannan 2004). Cur-
15
rent theory suggests that spatial structure facilitates coevolution by constraining
phenotypes in local populations that differ across larger geographical areas (Gibert
et al. 2013). Other theory has described how the spatial scales of patterns in host-
parasite coevolution are determined by spatial movement and the nature of the
coevolutionary interaction (Week and Bradburd 2023). An ambitious attempt to
describe the spatial dynamics of coevolution, called the geographic mosaic theory,
postulates that spatial structure and environmental heterogeneity creates coevo-
lutionary “hot” and “cold” spots in the arms race, leading to a mosaic of selection
pressures and a hypothesized dynamic process known as “trait remixing” (Thomp-
son 2005). Recent studies have tried to infer the contribution of coevolution to spe-
ciation and diversification (Hembry, Yoder, and Goodman 2014), but there remains
uncertainty about the degree to which coevolution is a primary mechanism for di-
versification of traits (Eaton 2008; Butler et al. 2009; Thompson 2009; Hembry, Yo-
der, and Goodman 2014; Parchman et al. 2016).
The well-studied predator-prey Taricha newt/Thamnophis garter snake system
is relatively well-understood and features an intriguing spatial mosaic of trait varia-
tion over a wide geographical range (Brodie, Feldman, et al. 2005; Hanifin, Brodie,
and Brodie 2008; Tseng 2011; Hague, Stokes, et al. 2020; Reimche et al. 2020). On
the Pacific Northwest coast of North America, various species of rough-skinned
newt in the genus Taricha often contain tetrodotoxin that poisons predators. One
of their predators, garter snakes in the genus Thamnophis, has developed resistance
to this toxin. Levels of toxicity and resistance are highly correlated across the re-
gion: areas where newt toxicity is high, snake resistance is usually also high; and
in areas where newt toxicity is low, snake resistance is usually also low. Further-
more, the snakes appear to be “winning” the coevolutionary arms race: snakes in
each area seem to be able to eat the local newts with relatively little ill effect, no
matter their toxicity (Hanifin, Brodie, and Brodie 2008; Feldman, Brodie Jr, et al.
2010).
These two striking observations – that the coevolutionary outcome varies strongly
across geography, and that the snakes nonetheless seem to be ahead of the newts –
provide additional clues about the underlying biological basis of the interaction.
Despite substantial work, the underlying cause of the observed geographic mosaic
of hotspots and coldspots is still unknown. Such mosaics of coevolving traits are
generally thought to be the result of heterogeneous ecological factors (e.g., resource
availability and/or differences in community composition), nonadaptive forces (e.g.,
16
local genetic drift and population structure), or both (Brodie, Ridenhour, and Brodie
2002; Hague, Stokes, et al. 2020; Thompson 2005). It would be of substantial inter-
est to know the balance of these forces in this particular case. To explain the obser-
vation that snakes tend to be “winning”, Feldman, Brodie Jr, et al. (2010) suggested
that the availability of large-effect resistance alleles allowed the snakes a “potential
‘escape’ from the arms race”. This leaves us with two questions: Is heterogeneity
in ecological factors required to explain the strikingly correlated maps presented in
Hanifin, Brodie, and Brodie (2008) and Hague, Stokes, et al. (2020), or can non-
adaptive forces lead to a mosaic? And, does a less-polygenic architecture provide
an advantage in this antagonistic coevolutionary arms race?
To answer these questions, as well as to better understand the dynamic in-
teraction between spatial structure, genetic architecture, and coevolution, we con-
ducted a simulation study, exploring a range of situations plausible for the Taricha
newt/Thamnophis garter snake system. In addition to answering these specific
questions, it is intriguing to consider other possible evolutionary outcomes, and
what conditions made this outcome possible. For instance, it is easy to imagine al-
ternate worlds in which snakes cannot eat newts (and survive), or in which snakes
only eat newts in locations where newts are less toxic. The many determinants of
the coevolutionary outcome include the strengths of various aspects of selection,
and the genetic opportunity for adaptation. The genetic basis of these traits in-
fluences how the traits vary within populations and how they respond to selective
pressures and environmental factors (Hoeksema and Forde 2008; Feldman, Brodie
Jr, et al. 2010), which can lead to different evolutionary outcomes. In snakes, there
are three known gene mutations that lead to high levels of tetrodotoxin resistance
(Feldman, Brodie, et al. 2012; McGlothlin et al. 2014). Each of these mutations al-
ters ion channel functioning, and so decreases the ability of a snake carrying the
mutation to crawl. The frequencies of each of these mutations vary over geographi-
cal space, and do not strongly correlate with local levels of newt toxicity (Geffeney
et al. 2005). On the other hand, the genetic basis of newt toxicity is still unknown.
A newt’s level of toxicity might be inducible and might be a result of environmen-
tal or bacterial factors. It is still unknown if it is heritable (Bucciarelli, Shaffer, et
al. 2017; Bucciarelli, Alsalek, et al. 2022; Vaelli et al. 2020). However, Bucciarelli,
Shaffer, et al. (2017) showed that newts retain a base level of toxicity even when
kept in lab conditions, and that young newts who are more toxic take longer to de-
velop.
17
To better understand the process of adaptive evolution across geographical
space, this paper asks three main questions: (1) How does the genetic architecture
of the traits in newts and snake affect how they coevolve? (2) Under what situa-
tions do we get spatial patterns of correlated traits in newts and snakes as we see
in the real world? (3) How fast does coevolution increase resistance and toxicity in
these organisms with different combinations of genetic architectures? In particular,
we compare different levels of genetic variance and polygenicity using individual-
based simulations of continuous geographic space. The results complement field ob-
servations by describing situations that are consistent with empirical observations,
and exploring other possible outcomes.
2.2 Methods
To explore these questions, we ran spatial individual-based, forward in time
simulations with SLiM (version 3, Haller and Messer 2019). The simulation had
two species that we call: “newts” and “snakes”, each with a quantitative trait for
toxicity and resistance, respectively. The simulated range was a large rectangular
map that spans 35 units in width and 140 units in length (long rectangular shape
similar to the west coast, larger simulations hampered the speed of the simulation).
The demographic details of this model were motivated by our current understand-
ing of the coevolutionary interactions of the rough-skinned newt (Taricha granu-
losa) and the garter snake (Thamnophis sirtalis). These simulations are a simplistic
representation of the system’s ecological and demographic complexities, but aim to
explore important aspects of the possible interaction dynamics.
The demographic model
Each simulated individual newt or snake was hermaphroditic and had a genome
of 108 base pairs, on which phenotype-affecting mutations occurred at a rate µ
base pair per generation. (Although real newts and snakes of these species are not
hermaphroditic, this should not significantly affect the dynamics). Each new muta-
tion had an “effect size” drawn from a Normal distribution with a mean of zero and
a standard deviation of σ (the values of µ and σ differed between simulations runs,
as described below). Demographic parameters were chosen that both newts and
snakes live for about 4 generations, on average. Newts and snakes had phenotypes
18
we call “toxicity” and “resistance”, respectively, that are determined by genotypes
in a way specified below. Each species also had fixed values for recombination rate
(10−8 crossovers per bp per generation) and for parameters controlling offspring dis-
persal and mate selection. The simulations used SLiM’s “non-Wright-Fisher” pop-
ulation model (Haller and Messer 2019) with overlapping generations and fluctu-
ating population sizes, with population dynamics described below. (We used SLiM
version 3.7.1 with different “populations” for each species; since then in version 4.0
direct support for multiple species was added (Haller and Messer 2023).)
Every generation, each newt finds a nearby newt with whom to mate and pro-
duce offspring. A newt at spatial location x with k neighbors at locations x1, . . . ,xk
would choose neighbor ias a mate with probability
∑ P (x− xi) , (2.1)k
j=1 P (x− xj)
where P () is the density of a 2-D Normal distribution with standard deviation of 1
unit. The number of offspring they produce is Poisson distributed with a mean of
1/4, and each offspring thus produced would disperse to a random location whose
displacement relative to the parent was drawn from a Normal distribution with a
mean of zero and a standard deviation of 1 unit. Local newt population density
was computed by smoothing using a Normal kernel: local density at location x is
∑k
r(x) = P (x− xj), (2.2)
j=1
where x1, . . . , xj are the locations of nearby newts (up to a maximum distance of 3
units).
Higher phenotypes were more costly than lower phenotypes. The probability
that a newt at location x survives to the next time step if they have phenotype z is
e−(z/c)
2 × 1min(0.95, ), (2.3)
1.2r(x)
where c is a parameter controlling the costliness of the phenotype (so that fitness
decreases as the phenotype increases). The probability of survival decreases with
density, due to competition between individuals of the same species. Consequently,
this leads to higher mortality in areas of higher densities. The parameters are cho-
sen so that the rough equilibrium density is one individual per unit area. Unless
otherwise stated, the value of c was set to 100.
19
Snake reproduction, offspring dispersal, and survival used the same dynamics
as the newts. Furthermore, at each time step, each nearby newt-snake pair could
“encounter” each other. To do this, we iterated over all snakes in random order;
choosing for each snake a set of nearby newts and resolving these encounters before
moving to the next snake. The probability that a snake “encounters” a given newt
at distance D is
−D
2
IR × e 2 , (2.4)
where IR is the baseline interaction rate for nearby individuals (set to 0.05 unless
otherwise noted), and that individuals that were closer were more likely to interact
than individuals further apart. When a snake and a newt encounter each other, the
outcome of the interaction depends on the difference between their phenotypes: if
L is the newts toxicity minus the snake’s resistance, then the probability that the
snake survives and the newt dies is
1
− . (2.5)1 + e L/10
The form is chosen so that phenotype differences must change by around 10 units
to substantially affect the interaction. If the snake does not eat the newt, the snake
dies and the newt survives. For a newt to have a good chance of surviving an in-
teraction with a snake, their level of toxicity needs to be greater than the snake’s
level of resistance. For each newt a snake consumes, the snake receives a “fitness
bonus” of 0.1, which is added to the probability of surviving to the next time step
(however, increases past 1.0 have no effect).
We used coalescent simulations produced by msprime (Kelleher, Etheridge, and
McVean 2016) to generate initial genetic variation. These simulations had an effec-
tive population size of 10,000 and a recombination rate of 10−8. Mutations’ effect
sizes were drawn from the same distribution as within SLiM, and added to the re-
sulting tree sequence using pyslim (Rodrigues, Ralph, et al. 2023) (these had no
effect on the coalescent simulation). To initialize each SLiM simulation, 300 indi-
viduals for each species were uniformly distributed across space.
Genetic Architecture
A major goal of our study is to describe how genetic architecture affects this
coevolving system. By “genetic architecture” we mean the combination of mutation
rate (µ) and the standard deviation of mutation effect size (σ, which we refer to
20
as “effect size”). We explored a range of values for mutation rate and effect size to
span from a highly polygenic to an oligogenic model (Table 1). Many evolutionary
dynamics depend primarily on mutational variance, which is
VM = µ× σ2. (2.6)
We model the phenotypes for each species as exponentials of additive genetic traits.
Concretely, if the sum of the effects of all mutations carried by an individual newt
is G, then that newt’s phenotype (i.e., toxicity) is eG/10. Snake resistance is deter-
mined in exactly the same way. The exponential transform is used here because
toxicity and resistance are non-negative, and because then mutations have multi-
plicative effects (i.e., increase or decrease the phenotypes by percentages).
Simulation Experiments
We ran four replicate simulations at each distinct set of parameter values,
across all experiments. Table 2.1 shows the parameters used in three experiments
that test how mutational variance impact the evolution of newt and snake pheno-
types through different combinations of mutation rate and mutation effect size. Our
first experiment set (Experiment 1) varies both mutation rates and mutation effect
sizes. In the second experiment, mutation rate was for both species in each simula-
tion, but allows the species to have different mutation effect sizes (and hence muta-
tional variance). In the third experiment, mutational variance (VM) was the same
for newts and snakes in each simulation, although polygenicity could be different
(varying mutation rate and mutation effect size).
Experiment 1 ran simulations with all possible combinations of the four ge-
netic architectures labeled “1a” to “4a” in Table 2.1. Since there are sixteen possi-
ble combinations (e.g., snakes had 1a and newts had 3a) and we ran four replicates
per combination, there was a total of 64 simulations. Across these genetic architec-
tures, mutation rate that ranged from 10−8 to 10−11, and the standard deviation of
mutation effect sizes (σ) ranged from 0.005 to 0.05. The genetic architectures are
arranged so that mutational variance increases with label (see rightmost column in
Table 2.1.
In Experiment 2, we study the effects of mutational variance. To do this, we
spanned the same range of parameters as in Experiment 1, in each simulation both
newts and snakes have the same mutation rate (µ), but have (possibly) different
21
mutation effect sizes (σ), and hence different mutational variances (VM = µσ2).
Note that in Table 2.1, genetic architectures with the same mutation rate are grouped
together in shaded groups of four rows, and that architectures in the same group
have the same numbered portion of their label. So, in each simulation as a part of
this experiment, newts and snakes were each assigned a genetic architecture from
the same shaded group: this led to sixteen combinations within each of the four
mutation rate groups, and hence 256 simulations across 64 distinct combinations of
genetic architectures.
In Experiment 3 we matched mutational variance between the species, and
looked for the effects of polygenicity. This was structured similarly to Experiment
2: shaded groups in Table 2.1 now have the same VM , and we ran simulations with
each of the sixty-four possible pairs of genetic architectures for which the two are
drawn from the same group (with replicates, 256 total simulations). For instance,
in one simulation, newts had genetic architecture with µ = 10−8 and σ = 0.0158
“1g” and snakes had genetic architecture with µ = 10−11 and σ = 0.5 “4g”; these
two have the same mutational variance (VM = 2.5 × 10−12), but different poly-
genicity (newts have a high rate of small mutations; snakes have a low rate of large
mutations). (Also note that the set of genetic architectures in Experiment 3 is the
same as in Experiment 2; what differs between the experiments is which pairs are
assigned to newts and snakes.)
The contribution of coevolution to phenotype change
To assess the strength of coevolution on phenotype change within the simula-
tions (as opposed to genetic drift), we ran additional simulations after modifying
the original model. We separately made two important changes, making either (1)
the trait (toxicity or resistance) not heritable, and (2) the outcome of the interac-
tion random (instead of dependent on phenotype). For (1), every new individual
had a phenotype not determined by their genetics but instead as eG/10 with G cho-
sen independently from a Normal distribution with mean set to 3 and a standard
deviation set at 2 (creating a phenotype near 1). For (2), we made the outcome of
each snake-newt interaction depend on the result of a fair coin flip instead of the
difference in phenotype: with 50% probability the snake eats a newt, otherwise the
snake dies.
To measure the speed of coevolution, we identified the “final mean phenotype”
22
Experiment Genetic Arch. Mutation Mutation Effect Mutational
(Newt/Snake) Rate (µ) Size (σ) Variance
1: Low to High 1a 10−8 0.005 (2V.5×) 10−13M
Mutational Variance 2a 10−9 0.05 2.5× 10−12
with changing 3a 10−10 0.5 2.5× 10−11
mutation rate 4a 10−11 5.0 2.5× 10−10
1b 10−8 0.005 2.5× 10−13
2b 10−8 0.0158 2.5× 10−12
3b 10−8 0.05 2.5× 10−11
4b 10−8 0.158 2.5× 10−10
1c 10−9 0.0158 2.5× 10−13
2: Low to High 2c 10−9 0.05 2.5× 10−12
Mutational Variance 3c 10−9 0.158 2.5× 10−11
(same mutation 4c 10−9 0.5 2.5× 10−10
rate) How does 1d 10−10 0.05 2.5× 10−13
increasing 2d 10−10 0.158 2.5× 10−12
mutational variance 3d 10−10 0.5 2.5× 10−11
4d 10−10 1.58 2.5× 10−10
1e 10−11 0.158 2.5× 10−13
2e 10−11 0.5 2.5× 10−12
3e 10−11 1.58 2.5× 10−11
4e 10−11 5.0 2.5× 10−10
1f 10−8 0.005 2.5× 10−13
2f 10−9 0.0158 2.5× 10−13
3f 10−10 0.05 2.5× 10−13
4f 10−11 0.158 2.5× 10−13
1g 10−8 0.0158 2.5× 10−12
3: Same Mutation 2g 10−9 0.05 2.5× 10−12
Variance How does 3g 10−10 0.158 2.5× 10−12
effects of 4g 10−11 0.5 2.5× 10−12
polygenicity effect 1h 10−8 0.05 2.5× 10−11
newt and snake 2h 10−9 0.158 2.5× 10−11
phenotype? 3h 10−10 0.5 2.5× 10−11
4h 10−11 1.58 2.5× 10−11
1i 10−8 0.158 2.5× 10−10
2i 10−9 0.5 2.5× 10−10
3i 10−10 1.58 2.5× 10−10
4i 10−11 5.0 2.5× 10−10
Table 2.1. A summary of all parameter sets used in simulations. Within each of the
three experiments, newts and snakes were assigned all combinations of genetic ar-
chitectures for which both rows are in the same group (colored blocks). Parameter
sets are labeled (second column) by an identifier parameter sets in the same group
have the same letter in their label.
23
as the average mean phenotype over the last 1000 time steps, and the “equilibrium
time” as the first time mean phenotype reached 98% of the final mean phenotype.
We then reported the speed as the difference in mean phenotype between the equi-
librium time and time step 100, divided by the number of elapsed time steps.
Heterogeneous landscapes
We also conducted simulations where some parameters changed across geo-
graphical space. In these simulations, we used the genetic architectures from Ex-
periment 1 (see Table 2.1) to see how newt and snake phenotypes changed when
the costliness of the phenotypes or the interaction rate varied across space. In each,
the relevant parameter varied across the map in a linear gradient. Each individual’s
location was then used to determine the correct value for that individual.
To simulate geographical variation in phenotype cost, the parameter c of equa-
tion (3.3) decreased linearly across the map, from c = 50 in the south to c = 250 in
the north. So, individuals that lived in the top portion of the map had a larger fit-
ness penalty for having a high phenotype than did individuals living in the bottom
portion of the map. We ran simulations in which there was geographical variation
in cost for both newts and snakes, as well as simulations in which cost varied for
only one species. The fitness cost ranged from 50 to 250 and impacted the proba-
bility of snake and newt survival (equation 3).
To simulate geographical variation in interaction rate, each snake’s location
was used to determine the interaction rate, i.e., the parameter IR from equation
(2.4), which is the base probability with which a snake would encounter a nearby
newt. This map had newts and snakes interacting more at the bottom of the map
than at the top of the map: IR ranged from 0.01 to 0.1 linearly with north-south
position.
Data collection
We used code in SLiM to collect data on newt and snake phenotypes across
the entire geographical area. In particular, we collected local newt and snake mean
phenotypes by using SLiM’s summarizeIndividuals() function to divide the map
up into a 5 × 20 grid and calculating the local mean phenotypes within each grid
24
cell. The local mean phenotypes were then used to calculate spatial correlations
between newt and snake phenotypes.
Data availability
The simulation and analysis code is accessible through github at https://github.
com/Vcaudill/Coevolution.
2.3 Results
Newt and snake Evolution
We first evaluated under which situations where phenotypes were actually coe-
volving. If coevolution was occurring, we anticipated an increase in newt and snake
average phenotypes over time. Figure 2.1 shows three common outcomes of the
simulation, with mean phenotypes for newts (red) and snakes (blue): fast evolution,
slow evolution, and no evolution. Simulation with relatively fast evolution had the
average newt and snake phenotype rising rapidly and reaching a steady-state point
before 10,000 generations, while relatively slow simulations took longer than 20,000
generations to equilibrate as “slow”. In simulations with no evolution, the average
phenotype of one or both species’ changed very little throughout the simulation.
Figure 2.2 shows the average change phenotypes (averaged across both species)
over the course of the simulation for all genetic architecture combinations from Ex-
periment 1. When the genetic architecture of either species had low mutational
variance (this was genetic architecture 1a, with lowest VM), there was no coevo-
lution: it’s mean phenotype did not increase (or decrease), regardless of the other
species’ genetic architecture. However, even when one species could not evolve, the
other species’ phenotype still might, due to genetic drift (e.g., snake phenotype in-
creased when newt phenotype could not increase in Figure 2.1C). The strongest
effect of coevolution – i.e., the largest change in both species’ phenotypes occurred
when both had intermediate mutational variance and polygenicity (genetic architec-
tures 2a and 3a). A species with the final genetic architecture (4a), which had the
largest amount of mutational variance (primarily from large-effect mutations) still
was able to coevolve, but showed a lower mean phenotype change.
It is interesting to note that the change in newt and snake mean phenotype
is not symmetrical with reciprocal genetic architecture combinations. For exam-
25
Figure 2.1. Mean phenotype dynamics over time in simulations showing three
speeds of evolution; fast, slow, and no. In the top panel (fast evolution), newt (red)
and snake (blue) average phenotypes go up quickly, reaching an equilibrium at
around 10,000 generations. In the middle panel (slow evolution), newt and snake
phenotypes rise more slowly, reaching equilibrium at around 20,000 generations. In
cases of “no evolution”, at least one species’ mean phenotype remains flat, in this
example (bottom panel), the mean newt phenotype remains flat.
26
Figure 2.2. Mean phenotypes for each combination of genetic architectures in Ex-
periment 1. Each box plot shows the range of mean phenotypes observed across
time (i.e., the entire simulation) and simulation replicate for one combination of
genetic architectures. The x-axis represented the genetic architecture of snake
(left) or newt (right), and color represents the genetic architecture of the opposite
species. Note that when either species has genetic architecture 1a (with lowest ge-
netic variance), that species has consistently low phenotype; otherwise, their mean
phenotype depends on the genetic architecture of the other species.
ple, when both newts and snakes had genetic architecture 2a the change in newt
phenotype was larger than the change in snake phenotype. There was also a rela-
tionship between phenotype and population size (see Supplementary Figures A.1
and A.2): species with higher average phenotype tended to have larger population
sizes, as one might expect if that species were “winning” the coevolutionary arms
race. (Even though we saw newt and snake population sizes fluctuate we did not
observe extinction.) In simulations without the lowest mutation variance (genetic
architecture 1a), newt and snake mean phenotypes seemed to be coevolving. How-
ever, this pattern could in principle be caused by drift instead of coevolution. To
exclude this possibility, we next modified key components of the simulation to ex-
amine what was driving the change in phenotype.
How much is coevolution driving phenotype change?
To verify that phenotypic changes were the result of coevolution, we altered
both heritability and (separately) the snake-newt interaction while keeping every-
27
thing else as similar as possible, so we could see what phenotypic changes were ex-
pected in the absence of coevolutionary forces (see Methods for details). If genetic
drift were responsible for the observed increase in phenotype, we would see similar
increases in phenotype in these simulations in which coevolution was not possible.
Figure 2.3 shows the average change in phenotype in these simulations, aver-
aged across both species, for different genetic architecture combinations across gen-
erations 100 to 50,000, compared to the previously described coevolutionary simu-
lations shown in Figure 2.1. For clarity, we do not show simulations in which either
species had low mutational variance (i.e., genetic architecture 1a). As expected,
we saw an increase in newt and snake phenotype only in the standard simulations
(orange boxplots), and not when either heritability or dependence of the interac-
tion on phenotype was removed (other boxes). In simulations where the interaction
does not depend on phenotype (green boxplots), phenotypes decreased. In these,
the phenotype provided no benefit, so the mean phenotype dropped to a level de-
termined by mutation-selection balance (where “selection” is due to the costliness
of the phenotype). Simulations in which the trait was not heritable (barely visible,
grey boxplots) had no change in newt and snake phenotype for all genetic architec-
tures, as expected. (However, these simulations still provide a meaningful control
from levels of spatial correlations expected in the absence of coevolution.)
Spatially Heterogeneous Landscapes
In the real world, newt and snake resistance are correlated across a broad ge-
ographical region (Hanifin, Brodie, and Brodie 2008; Reimche et al. 2020) in loca-
tions where newts are very toxic, snakes tend to be very resistant to their toxin,
and vice-versa. The empirical (product-moment) correlation between the toxicity
and resistance values reported in Hanifin, Brodie, and Brodie (2008) is r = 0.77.
Do our simulations recapitulate this spatial correlation?
The answer, so far, is “no”: Figure 2.4 shows that no genetic architecture com-
bination had a comparable degree of spatial correlation. (To obtain a roughly equiv-
alent measure from our simulations, we split the entire area into 100 smaller squares
(in a 5 × 20 grid), calculated local mean phenotypes of each species in the squares
2.5, and computed the correlation between these two vectors of local mean phe-
notypes; see Figure 2.5 for an example.) Although local fluctuations driven by co-
evolution could in principle have created spatial correlations (Thompson 2005), it
28
Figure 2.3. Newt and snake mean phenotypes (averaged together), comparing
standard simulations to simulations without heritability or random interaction out-
comes (see Methods). In the standard simulations (orange) the mean phenotype
for newts and snakes increases. When there is no heritability (grey), phenotypes
remain close to zero. When the newt-snake interaction outcome is random (not
based on phenotype) the average newt and snake phenotypes decreased (green).
appears that such fluctuations do not appear strongly, at least across this range of
parameter values in a homogeneous landscape.
However, the landscape inhabited by the real newts and snakes is not constant.
It seems very plausible that the real environment is heterogeneous – for instance,
(Reimche et al. 2020) found that an elevation gradient in the Sierra Nevadas was
correlated to levels of toxicity and resistance in a sister species of newts and snakes.
Furthermore, variation in population density, predator pressure, and/or habitat
could easily lead to varying rates of interaction between newts and snakes across
the landscape. So, we also ran simulations in heterogeneous landscapes. We changed,
in separate simulations, the costliness of the phenotype and the rate of interaction
between newts and snakes across the map. When the costliness of the phenotype
varied across space, phenotypes were even more strongly correlated than in the
29
Figure 2.4. Spatial newt and snake phenotype correlations for all genetic architec-
ture combinations, for three types of map. In each plot, the leftmost (green) box-
plot displays correlations across all combinations of genetic architectures containing
the lowest genetic variance (1a); other boxes (blue) show the range of spatial cor-
relations across replicates and time steps. The dashed line shows the empirical
newt and snake spatial correlation (reported by Hanifin, Brodie, and Brodie 2008).
When there is no spatial heterogeneity in the simulation (top), there is little spa-
tial correlation. Higher spatial correlations occur when newts and snake coevolve
on a map with heterogeneity in phenotype costliness (middle) or interaction rate
(bottom). Correlations are similar across all genetic architecture combinations with
high enough mutational variance.
empirical data (mean r = 0.84, median r = 0.90, empirical r = 0.77). To make
sure that this correlation caused by coevolution and not changing cost alone, we
re-ran this simulation without a phenotype-based interaction (as described above)
and saw no correlations (see Supplementary Figure A.3). Simulations in which only
one species’ costliness varied across the map had lower but still strong correlations
(snakes vary: mean r = 0.47; median r = 0.49; newts vary: mean r = 0.57, median
r = 0.70; see Supplementary Figure A.3). In addition to having high correlations
between phenotypes, all simulations with varying parameters across the map had a
wide range of phenotypic values, as seen in the empirical data (as in Figure 2.4).
We also saw high correlations in simulations where the interaction rate varied
30
Figure 2.5. Local newt and snake mean phenotypes across the geographical area at
generation 44501 in a simulation in which costliness of the phenotype was a spatial
gradient (higher costliness at the top of the map). The genetic architectures used
were snake: 4a and newt: 3a; other combinations are roughly similar. Each box
shows the mean phenotype of individuals living within that grid cell. Brighter col-
ors represent higher levels of toxicity or resistance.
31
across the map (mean r = 0.60). Varying interaction rate led to instances where
local extinction and subsequent recolonizations occurred in small sections of the
map. Adjusting the range of interaction rates or perhaps other parameters would
likely increase correlations, but we did not explore parameter space further.
Effects of Genetic Architecture on Coevolution
In Figure 2.2, we saw that the coevolutionary equilibrium values of newt and
snake mean phenotypes differed substantially depending on the genetic architec-
ture (and, interestingly, in an asymmetric way). However, we saw in Figure 2.4 that
genetic architecture did not affect the spatial correlation of newt and snake pheno-
type. How else does genetic architecture affect the coevolutionary dynamics? And,
does it affect which species is “winning”? Our simulations in Experiment 1 had ge-
netic architectures that simultaneously varied mutational variance and polygenic-
ity, so to disentangle these two effects, we ran two additional sets of simulations in
which we (Experiment 2) constrained newts and snakes to have the same mutation
rate (µ) but differing effect size (σ) and hence differing mutational variance (VM
from equation (2.6)); and (Experiment 3) constrained newts and snakes to have the
same mutational variance but different combinations of mutation rate and effect
size.
We saw that the speed of coevolution depends more on mutational variance
than it does on polygenicity for both newts (red) and snakes (blue) (Figure 2.6).
When mutational variance was high the speed of coevolution was fast, essentially
taking less time for the simulation to reach a steady-state. When mutational vari-
ance was lower, it took longer for the simulation to reach a steady-state.
Does it make sense to say that one species is “winning”? If one species has a
higher mean phenotype than the other, then that species will more often “win”
in encounters between the two species. Figure 2.7 shows mean phenotype differ-
ences between the species across the genetic architecture combinations of Experi-
ment 1 (excluding those combinations with architecture 1a, as usual). Here, we see
that population size differences and mean phenotype differences track each other
(also see in Supplementary Figure A.2). We also see that newts more often have
the higher mean phenotype – apparently, the asymmetry in their ecological inter-
action gives newts the coevolutionary edge. However, in at least one situation of
Figure 2.7, the snakes (on average) had a larger mean phenotype and population
32
Coevolution Slope
2d/2d         2d/3d         3d/2d         3d/3d             2h /2h          2h/3h         3h/2h         3h/3h
2c/2c         2c/3c          3c/2c          3c/3c              2g /2g          2g/3g         3g/2g         3g/3g
Genetic Architectures (red=newt, blue=snake)
Figure 2.6. The speed of coevolution depends on mutational variance and not
just the mutation rate or mutation effect size. Each boxplot shows the range of
coevolution speed (see Methods) across simulation replicates at that combination
of genetic architectures. Left plots show combinations in which mutation rate is
matched between species, arranged so that mutational variance increases mov-
ing left to right. Right plots show combinations in which mutational variance is
matched between species, arranged so that effect size (σ) increases and mutation
rate (µ) decreases moving left to right. Note that the combination 3d/3d has the
same amount of genetic variation as all combinations in the top-right panel (2h/2h,
2h/3h, 3h/2h, and 3h/3h) and that the combination 2c/2c has the same amount of
genetic variation as those in the bottom-right panel.
33
Coevolution Speed (Phenotype Change/Time)
size.
Supplementary Figure A.4 shows similar plots for Experiment 2, where the
two species have different mutational variances but the same mutation rate, across
four distinct mutation rates. In these simulations, although equilibrium pheno-
types depended on the genetic architectures, in general the species with the larger
mutational variance usually did worse, i.e., had a lower phenotype than the other
species. (This is seen in Supplementary Figure A.4 by the purple box plots moving
down to the right.) However, when the two species had the same mutational vari-
ance, newts usually had higher phenotype than snakes.
On the other hand, Figure A.5 shows that if both species have the same muta-
tional variance, then differences in polygenicity do not have a strong effect on the
outcome. (There are perhaps relatively minor differences in the equilibrium, but
patterns are unclear.) In summary, we have good evidence for a strong effect of mu-
tational variance, but not polygenicity, on which species is winning the coevolution-
ary arms race.
2.4 Discussion
In summary, we found that spatial heterogeneity in ecological factors was im-
portant for creating a spatial mosaic of correlated phenotypes. On the other hand,
our simulations could not produce such coevolutionary mosaics in the absence of
spatial heterogeneity, regardless of the genetic architecture. The genetic architec-
ture did affect the dynamics of coevolution: the speed of coevolution depended pri-
marily on mutational variance (in particular, if mutational variance was too low,
the species would not coevolve). However, genetic architectures with the same mu-
tational variance but different mutation rates and polygenicities had nearly identi-
cal coevolutionary dynamics.
One of our main goals was to identify conditions that would create the strik-
ing spatial correlations seen between the two species’ phenotypes in the wild. Even
though our simulations had enough space for local adaptation to occur, we saw no
such geographical correlations in simulations on uniform landscapes, regardless of
genetic architecture. However, environmental gradients in various aspects of the un-
derlying biology were sufficient to create geographical correlations of similar magni-
tude to that seen in the wild. This result, again, did not depend strongly on genetic
architecture.
34
Figure 2.7. Distribution of (A) mean newt and snake phenotypes, and (B) newt
and snake population sizes, and their differences (snake minus newt), from a reg-
ularly spaced set of generations between 45,000 and 50,000. All combinations of
genetic architecture from Experiment 1 are shown, except those containing the low-
est mutational variance (1a). The combination of genetic architectures is shown on
the x-axis labels, prepended with the difference of snake and newt log10(VM) values:
for instance, the leftmost set of boxplots, labeled “(-2)S:2a/N:4a”, refers to simula-
tions in which snakes have genetic architecture 2a, newts have genetic architecture
4a; and the snake’s genetic architecture has 100 times less mutational variance than
does the newt’s.
35
Genetic architecture affected the dynamics of coevolution: species with larger
mutational variance could increase their phenotype faster (and so “win”, at least
transiently). Furthermore, effects of genetic architecture on the final steady state
seemed mostly attributable to mutational variance – genetic architectures with the
same mutational variance but different polygenicities did not show large differences,
at least over the fairly coarse scale we examined. Our results (e.g., Figure 2.7) sug-
gest that there are indeed differences in the steady state achieved by different com-
binations of genetic architectures, but these differences are usually smaller than the
generation-to-generation noise observed over the course of our simulations.
Relationship to Other Models of Coevolution
Many previous theoretical papers also study coevolutionary dynamics. For
instance, Nuismer, Thompson, and Gomulkiewicz (2000) study how coevolution
might lead to clines in allele frequency using deterministic models in either discrete
or continuous space in which each species carries a single biallelic locus that medi-
ates the coevolutionary dynamics. Nuismer, Thompson, and Gomulkiewicz (2000)
found many situations in which coevolution maintained spatial variation in allele
frequencies, although their model did not include escalatory (i.e., arms race) an-
tagonism. Nuismer, Ridenhour, and Oswald (2007) develop a quantiative model of
trait coevolution in which – much like in our own – quatitative traits are additive
subject to stabilizing selection, and the result of antagonistic inter-species interac-
tions is determined by a logistic function of the difference in trait values (so that
each species “tries to exceed” the other). Nuismer, Ridenhour, and Oswald (2007)
deal with genetic architecture in three ways: by assuming fixed genetic variance
and using a quantitative genetics model; by assuming weak selection and quasi-
linkage equilibrium; and by doing deterministic numerical simulations in which ge-
netic variance is due to a fixed number of explicitly represented loci evolving under
strong selection. Nuismer, Gomulkiewicz, and Ridenhour (2010) also used a quan-
titative genetics model (i.e., of many small-effect loci but without explicitly repre-
senting the loci) in both analytical calculations and individual-based simulations of
an island model. Nuismer, Gomulkiewicz, and Ridenhour (2010) found that condi-
tions under which traits were correlated across space differed between across types
of coevolutionary interactions. They also found differences between the analytical
predictions and individual-based simulation results.
36
Our study looks at many of the same questions, using some of the same quan-
titative genetics tools. However, many aspects of models that are fixed in previous
work (e.g., number of polymorphic loci or population size) are emergent properties
of our simulations. We focus on using individual-based simulations to test how ge-
netic architectures affect coevolutionary trajectories with explicit models of contin-
uous space and genetic inheritance. It would not be possible to address some of our
questions using previous models: for instance, we study a broad range of genetic ar-
chitecture polygenicities, which would not be possible under a purely quantitative
genetics model. Furthermore, our model explicitly represents ecological dynamics –
so, for instance, local extinction/recolonization is possible, unlike under the fixed-
population-density models of many previous papers. Additional points of realism
of our simulations that make analysis of mathematical models difficult include re-
combination, fluctuating population sizes, and stochasticity. These are relatively
straightforward to implement thanks to the simulation software, SLiM, especially in
its’ newest release with improved support for interacting species (Haller and Messer
2023).
What Does This Tell us About the Newts and Snakes?
Our results support the idea that the spatial patterns in toxicity and resis-
tance observed in Taricha newts and Thamnophis garter snakes is a result of spa-
tial heterogeneity in some ecologically important parameters, rather than simply
spatial structure and resulting decoupling of local dynamics. It has been observed,
for instance, that toxicity of one Taricha species is correlated with elevation in the
Sierra Nevada (Reimche et al. 2020). There are many possible aspects that might
vary across space; variation in both trait costliness and interaction rate had simi-
lar effects. Hanifin, Brodie, and Brodie (2008) presented data across a large area of
the Pacific Northwest coast of North America, from Canada to Southern Califor-
nia, which spans a wide range of temperature, rainfall, and biodiversity. It is easy
to imagine that, for instance, the costliness to toxin-resistant snakes of being less
able to crawl quickly might vary with temperature, or that varying density of other
species (e.g., toads or owls that prey on newts) might lead to spatial patterns in
interaction rates (Toju and Sota 2006; Craig, Itami, and Horner 2007).
Feldman, Brodie Jr, et al. (2010) hypothesized that snakes were “winning” the
coevolutionary arms-race against newts due to the snakes’ genes of large effect. We
37
found that mutational variance, rather than polygenicity, was the important deter-
minant of which species was ahead (Figure 2.7). In fact, we found that in our sim-
ulations, newts – not snakes – had higher average phenotypes across most combi-
nations of genetic architectures, although the effect was relatively small and some-
what hard to predict. Other scenarios are possible that would explain this situa-
tion: for instance, perhaps high phenotype values are more costly for newts than
for snakes (in our simulations, we might set c smaller for newts than for snakes).
Or, perhaps there are relatively hard constraints on the upper limit of tetrodotoxin
production by newts, thus limiting the evolvability of the newt’s phenotype (which
is particularly plausible if the toxin is produced by an environmentally-acquired
bacteria (Vaelli et al. 2020)).
In summary, what have we learned about the newts and the snakes? Two ma-
jor takeaways are that (a) the observed spatial correlations in phenotypes is en-
tirely consistent with the proposed evolutionary story developed in the empirical
literature, but that (b) existence of these spatial patterns does not strongly con-
strain the set of possible underlying genetic architectures.
Spatial Patterns and Interactions
When evolution occurs across continuous space, evolutionary change and differ-
entiation might lead to differences in different locations. However, we saw no such
spatial patterns in simulations with a uniform environment. This might be due to
the mixing action of migration, but even across a very large range, we still might
not see substantial spatial patterns if evolution towards the same equilibrium was
occurring at the same rate in all locations. (If ecological factors are heterogeneous,
as in Figure 2.5, the equilibrium may differ across space.) In the real world, spatial
patterns are not ubiquitous: for instance, Hoeksema and Forde (2008) did not find
a relationship between spatial scale and degree of local adaptation.
Spatial patterns in biological systems are possible even without variation in the
underlying environment. Even on a scale of meters, spatial patterns can sponta-
neously emerge from ecological dynamics, e.g., of vegetation in grasslands (Thomp-
son and Daniels 2010; Sasaki 1997). Spatial cycling in reaction-diffusion equations
is known to produce striking spatial patterns in certain circumstances (e.g., Mur-
ray 1982; Britton 1990; Schreiber and Killingback 2013). It could be, for instance,
that local extinction-recolonization dynamics might play a role in the observed spa-
38
tial patterns. This might work, for instance, by cycles of (a) newts evolve toxic-
ity; (b) snakes evolve resistance; (c) snakes eat most of the newts; (d) snakes lose
resistance; (e) newts recolonize from a nearby relatively snake-free environment.
Alternatively, if snakes have an “all-or-nothing” genetic architecture of resistance
and newts have relatively little genetic variance for toxicity at any time, then one
might imagine a cycle of: (a) newts evolve toxicity; (b) snakes evolve resistance; (c)
snakes are sufficiently resistant that variation in newt toxicity levels provides no
advantage; (d) newts lose costly toxicity; (e) snakes lose resistance. Across space,
these cycles could be decoupled between regions, and barriers to dispersal would
make it more likely they are not synchronized, leading to spatial patterns. How-
ever, there is no evidence that such cycles occur in the real world. In this paper,
we have not attempted to search for conditions that would produce such cycling;
instead, our goal was to see whether, under reasonably plausible conditions, such
dynamics might occur. It would, however, be interesting to investigate.
Our simulations highlighted the importance of understanding coevolutionary
systems in a broader context: in our simulations, newts and snakes evolved in re-
sponse not only to the selective pressure of coevolution, but also the heterogeneous
environment. This point was also made by Yoder and Nuismer (2010), who sug-
gested that heterogeneity in the environment was also necessary for coevolution-
ary dynamics to increase phenotypic diversity. Furthermore, similar coevolutionary
mosaics could be produced by distinct genetic and environmental conditions, and
so care must be taken in ascribing specific underlying mechanisms to patterns ob-
served in the world (Craig and Itami 2021). This work highlights the importance
of considering things outside of the focal species interaction (e.g., indirect interac-
tions).
Genetic Architecture
The underlying genetic mechanism of a trait determines how a trait can evolve
and has implications on diversification and genetic variation (Hague, Feldman,
et al. 2017). In our simulations, a species with sufficiently low mutational vari-
ance would never adapt, due to a lack of genetic variation. (For instance, if mu-
tational variance for snakes was sufficiently low and newts were somewhat toxic,
then rare resistance mutations in snakes might not increase snake phenotypes suf-
ficiently on their own to meaningfully increase snake fitness.) Lack of genetic vari-
39
ation is thought to be a limiting factor in some potentially coevolutionary interac-
tions (Hoeksema and Forde 2008).
Quantitative genetics models can predict how traits might change due to selec-
tive pressures (Lynch and Walsh 1998). These models typically depend only on ge-
netic variance, without specifying the polygenicity of the underlying genetic archi-
tecture. We ran simulations under various architectures that ranged from simple (a
few mutation of large effect) to complex (many mutations of minor effect), and al-
though mutational variance indeed was the primary determining factor, polygenic-
ity did affect steady state of the phenotype, although in complex ways. It would be
interesting to model how deviations from the highly polygenic limit described by
quantitative genetics affects results in practice.
Consistent with predictions of quantitative genetics, the speed of coevolution
was mediated by the level of mutational variance created by the genetic architec-
ture, rather than mutation rate or mutation effect size separately. The species that
had a higher mutational variance often evolved and reached a steady state quicker.
However, there were instances were the highest mutational variance limited a species’
ability to coevolve. This was potentially due to large effect mutations interacting
with the costliness of the phenotype: a particularly large effect mutation might
cause an individual to have a phenotype so high that it perished, thus slowing the
speed of coevolution.
Perhaps counterintuitively, we found that higher mutational variance seemed
to put a species at a coevolutionary disadvantage – see e.g., Figure 2.7. To un-
derstand this, consider what is happening in the simulations represented by the
furthest-left boxplots of that figure. Here, newts have much higher mutational vari-
ance than snakes, and hence a much wider distribution of phenotypes: if this is
sufficiently wide, then even if the mean newt phenotype is higher than the mean
snake phenotype, there will still be many newts that can be eaten by the typical
snake. Furthermore, snakes have a much higher population size than newts, so a
given newt is more likely to encounter a snake than vice-versa, and so the selection
pressure on newts to have a high phenotype is stronger than on snakes. A similar
argument applies to the other end of the parameter range.
40
Limitations and Continuing Questions
Our simulation was inspired by ecological interactions of real newts and snakes,
but has a great many simplifications. For instance, newt and snake demographies
(e.g., dispersal mechanism, birth rates, death rate, and age structure) were identi-
cal, despite Taricha newts living substantially longer (on average) than Thamnophis
garter snakes in the wild. These aspects of life history impact the evolutionary tra-
jectories of traits. Furthermore, many aspects of newt toxicity are unknown: it is
thought that newt’s tetrodotoxin may be produced by a bacterium (as do other
tetrodotoxin-producing species, Vaelli et al. 2020; Bucciarelli, Alsalek, et al. 2022),
but whether any such bacterium would be acquired from the environment or ver-
tically transmitted is unknown. However, it seems likely that there would still be
genetic variation related to the production and/or tolerance of the toxin.
In our simulation, if an interaction occurs between a newt and a snake, one
perishes while the other survives. In the wild, a snake may attempt to eat a newt,
fail, but both species survive (Reimche et al. 2020). Such interactions where both
survive may lead to different results if trait plasticity is considered. It has been
shown that a newt’s toxicity can increase and remain high for some time after the
newt is be disturbed (Bucciarelli, Shaffer, et al. 2017). Therefore, a failed attack
could leave a newt in a better position if more snakes were nearby. Relatively lit-
tle is known about how often snakes and newts interact in the wild, and how much
that varies across space or time.
Finally, it would be interesting to incorporate more specifics about the known
genetic architecture of tetrodotoxin resistance. We did not pursue this direction in
this paper because we were interested not only in conclusions about the real sys-
tem, but also broader evolutionary questions about the role of genetic architecture
in these sorts of coevolutionary dynamics. We have aimed more generally to cap-
ture important aspects of the dynamics without being distracted by unnecessary
details. Modeling, particularly simulation-based exploration, walks a fine line be-
tween specificity (to faithfully represent an aspect of reality) and generality (so that
results are generalizable). However, future work in this area could reflect new un-
derstandings of the system, as well as more realistically model differences between
the species such as genome sizes and life cycles.
41
2.5 Bridge
The results from this chapter highlighted the importance of the various mech-
anisms that impact evolutionary trajectories. In this chapter, evolution exhibited
stochastic behavior, but was influenced by a range of parameters used to model a
specific evolutionary system. In the next chapter I explore the real-world outcomes
of evolution by examining the signals left in the genomes of 14 cichlid species. I will
investigate how different mechanisms of evolution have impacted their genetic land-
scapes
2.6 References
Berenbaum, M. R., A. R. Zangerl, and J. K. Nitao, Constraints on chemical coevo-
lution: Wild parsnips and the parsnip webworm (en), Evolution 40, no. 6, 1215–
1228, doi: 10.1111/j.1558-5646.1986.tb05746.x.
Bohannan, B.J.M. and R.E. Lenski, Linking genetic change to community evolu-
tion: insights from studies of bacteria and bacteriophage (en), Ecology Letters 3,
no. 4, 362–377, doi: 10.1046/j.1461-0248.2000.00161.x.
Britton, N. F., Spatial Structures and Periodic Travelling Waves in an Integro-
Differential Reaction-Diffusion Population Model, SIAM J. Appl. Math. 50,
no. 6, 1663–1688, doi: 10.1137/0150099.
Brodie, E. D., Reciprocal Selection at the Phenotypic Interface of Coevolution (en),
Integrative and Comparative Biology 43, no. 3, 408–418, doi: 10.1093/icb/43.3.
408.
Brodie, Edmund D Jr, B J Ridenhour, and Edmund D III Brodie, THE EVOLU-
TIONARY RESPONSE OF PREDATORS TO DANGEROUS PREY: HOTSPOTS
AND COLDSPOTS IN THE GEOGRAPHIC MOSAIC OF COEVOLUTION
BETWEEN GARTER SNAKES AND NEWTS (en), Evolution; international
journal of organic evolution 56(10), 2067–2082.
Brodie, Edmund D., Chris R. Feldman, et al., Parallel Arms Races between Garter
Snakes and Newts Involving Tetrodotoxin as the Phenotypic Interface of Co-
evolution (en), Journal of Chemical Ecology 31, no. 2, 343–356, doi: 10.1007/
s10886-005-1345-x.
42
Bucciarelli, G.M., Farid Alsalek, et al., Toxic Relationships and Arms-Race Coevo-
lution Revisited (en), Annual Review of Animal Biosciences 10, no. 1, 63–80,
doi: 10.1146/annurev-animal-013120-024716.
Bucciarelli, Gary M., H. Bradley Shaffer, et al., An amphibian chemical defense
phenotype is inducible across life history stages (en), Scientific Reports 7, no. 1,
8185, doi: 10.1038/s41598-017-08154-z.
Butler, Richard J. et al., Testing co-evolutionary hypotheses over geological timescales:
interactions between Mesozoic non-avian dinosaurs and cycads (en), Biological
Reviews 84, no. 1, 73–89, doi: 10.1111/j.1469-185X.2008.00065.x.
Craig, Timothy P. and Joanne K. Itami, A geographic mosaic of coevolution be-
tween Eurosta solidaginis (Fitch) and its host plant tall goldenrod Solidago al-
tissima (L.) (en), Evolution 75, no. 12, 3056–3070, doi: 10.1111/evo.14391.
Craig, Timothy P., Joanne K. Itami, and John D. Horner, Geographic variation in
the evolution and coevolution of a tritrophic interaction: geographic coevolution
in Eurosta solidaginis (en), Evolution 61, no. 5, 1137–1152, doi: 10 . 1111/ j .
1558-5646.2007.00099.x.
Daugherty, Matthew D. and Harmit S. Malik, Rules of Engagement: Molecular In-
sights from Host-Virus Arms Races (en), Annual Review of Genetics 46, no. 1,
677–700, doi: 10.1146/annurev-genet-110711-155522.
Dodds, Peter N. et al., Direct protein interaction underlies gene-for-gene specificity
and coevolution of the flax resistance genes and flax rust avirulence genes (en),
Proceedings of the National Academy of Sciences 103, no. 23, 8888–8893, doi:
10.1073/pnas.0602577103.
Eaton, C.D., Coevolutionary Research (en), Encyclopedia of Ecology, Elsevier, 659–
663, doi: 10.1016/B978-008045405-4.00777-1.
Ehrlich, Paul R and Peter H Raven, Butterflies and Plants: A Study in Coevolution
(en), Evolution 18, no. 4, 586–608.
Feldman, Chris R, Edmund D Brodie Jr, et al., Genetic architecture of a feeding
adaptation: garter snake ( Thamnophis ) resistance to tetrodotoxin bearing prey
(en), Proceedings of the Royal Society B: Biological Sciences 277, no. 1698,
3317–3325, doi: 10.1098/rspb.2010.0748.
43
Feldman, Chris R., Edmund D. Brodie, et al., Constraint shapes convergence in
tetrodotoxin-resistant sodium channels of snakes (en), Proceedings of the Na-
tional Academy of Sciences 109, no. 12, 4556–4561, doi: 10.1073/pnas.1113468109.
Forde, Samantha E., John N. Thompson, and Brendan J. M. Bohannan, Adaptation
varies through space and time in a coevolving host–parasitoid interaction (en),
Nature 431, no. 7010, 841–844, doi: 10.1038/nature02906.
Geffeney, Shana L. et al., Evolutionary diversification of TTX-resistant sodium
channels in a predator–prey interaction (en), Nature 434, no. 7034, 759–763,
doi: 10.1038/nature03444.
Gibert, Jean P. et al., The Spatial Structure of Antagonistic Species Affects Coevo-
lution in Predictable Ways, The American Naturalist 182, no. 5, 578–591, doi:
10.1086/673257.
Hague, Michael T. J., Amber N. Stokes, et al., The geographic mosaic of arms race
coevolution is closely matched to prey population structure (en), Evolution Let-
ters 4, no. 4, 317–332, doi: 10.1002/evl3.184.
Hague, Michael T.J., Chris R. Feldman, et al., Convergent adaptation to dangerous
prey proceeds through the same first-step mutation in the garter snake Thamnophis
sirtalis (en), Evolution 71, no. 6, 1504–1518, doi: 10.1111/evo.13244.
Haller, Benjamin C and Philipp W Messer, SLiM 3: Forward Genetic Simulations
Beyond the Wright–Fisher Model (en), Molecular Biology and Evolution 36,
no. 3, 632–637, doi: 10.1093/molbev/msy228.
, SLiM 4: multispecies eco-evolutionary modeling, The American Naturalist
201, no. 5, E127–E139.
Hanifin, Charles T, Edmund D Jr. Brodie, and Edmund D III Brodie, Phenotypic
Mismatches Reveal Escape from Arms-Race Coevolution (en), PLoS Biology 6,
no. 3, e60, doi: 10.1371/journal.pbio.0060060.
Hembry, David H., Jeremy B. Yoder, and Kari Roesch Goodman, Coevolution and
the Diversification of Life (en), The American Naturalist 184, no. 4, 425–438,
doi: 10.1086/677928.
Hoeksema, Jason D. and Samantha E. Forde, A Meta-Analysis of Factors Affect-
ing Local Adaptation between Interacting Species (en), The American Naturalist
171, no. 3, 275–290, doi: 10.1086/527496.
44
Janzen, Daniel H, Coevolution of Mutualism Between Ants and Acacias in Central
America (en), Evolution 20, no. 3, 249–275.
Janzen, Daniel H et al., When is it coevolution, Evolution 34, no. 3, 611–612.
Kelleher, Jerome, Alison M Etheridge, and Gilean McVean, Efficient Coalescent
Simulation and Genealogical Analysis for Large Sample Sizes, PLOS Computa-
tional Biology 12, no. 5, 1–22, doi: 10.1371/journal.pcbi.1004842.
Lynch, Michael and Bruce Walsh, Genetics and analysis of quantitative traits, 1.
McGlothlin, Joel W. et al., Parallel Evolution of Tetrodotoxin Resistance in Three
Voltage-Gated Sodium Channel Genes in the Garter Snake Thamnophis sirtalis,
Molecular Biology and Evolution 31, no. 11, 2836–2846, doi: 10.1093/molbev/
msu237, eprint: https://academic.oup.com/mbe/article- pdf/31/11/2836/
13169900/msu237.pdf.
Murray, J D, Parameter space for Turing instability in reaction diffusion mecha-
nisms: a comparison of models, J Theor Biol 98, no. 1, 143–163.
Nuismer, Scott L., Richard Gomulkiewicz, and Benjamin J. Ridenhour, When Is
Correlation Coevolution? (en), The American Naturalist 175, no. 5, 525–537,
doi: 10.1086/651591.
Nuismer, Scott L., Benjamin J. Ridenhour, and Benjamin P. Oswald, ANTAGO-
NISTIC COEVOLUTION MEDIATED BY PHENOTYPIC DIFFERENCES
BETWEEN QUANTITATIVE TRAITS (en), Evolution 61, no. 8, 1823–1834,
doi: 10.1111/j.1558-5646.2007.00158.x.
Nuismer, Scott L., John N. Thompson, and Richard Gomulkiewicz, COEVOLU-
TIONARY CLINES ACROSS SELECTION MOSAICS, Evolution 54, no. 4,
1102–1115.
Parchman, Thomas L. et al., Genome divergence and diversification within a geo-
graphic mosaic of coevolution (en), Molecular Ecology 25, no. 22, 5705–5718,
doi: 10.1111/mec.13825.
Reimche, Jessica S. et al., The geographic mosaic in parallel: Matching patterns of
newt tetrodotoxin levels and snake resistance in multiple predator–prey pairs
(en), Journal of Animal Ecology 89, no. 7, 1645–1657, doi: 10 . 1111 / 1365 -
2656.13212.
45
Rodrigues, Murillo F., Peter L. Ralph, et al., PySLiM manual, url: https://tskit.
dev/pyslim/docs/latest.
Sasaki, Akira, Clumped Distribution by Neighbourhood Competition, J. Theoret.
Biol. 186, no. 4, 415–430, doi: 10.1006/jtbi.1996.0370.
Schreiber, Sebastian J. and Timothy P. Killingback, Spatial heterogeneity promotes
coexistence of rock–paper–scissors metacommunities, Theoretical Population Bi-
ology 86, 1–11, doi: https://doi.org/10.1016/j.tpb.2013.02.004.
Thompson, J.N., The Geographic Mosaic of Coevolution, Interspecific Interactions,
University of Chicago Press.
Thompson, John N., The Coevolving Web of Life (American Society of Naturalists
Presidential Address) (en), The American Naturalist 173, no. 2, 125–140, doi:
10.1086/595752.
Thompson, Sally E. and Karen E. Daniels, A Porous Convection Model for Grass
Patterns. The American Naturalist 175, no. 1, E10–E15, doi: 10.1086/648603,
eprint: https://doi.org/10.1086/648603, PMID: 19911904.
Thrall, Peter H. et al., Coevolution of symbiotic mutualists and parasites in a com-
munity context, Trends in Ecology and Evolution 22, no. 3, 120–126, doi: 10 .
1016/j.tree.2006.11.007.
Toju, Hirokazu and Teiji Sota, Imbalance of Predator and Prey Armament: Geo-
graphic Clines in Phenotypic Interface and Natural Selection. (en), The Ameri-
can naturalist, 167(1), doi: 10.1086/498277.
Tseng, Michelle, In the light of evolution: essays from the laboratory and field.
Vaelli, Patric M et al., The skin microbiome facilitates adaptive tetrodotoxin produc-
tion in poisonous newts (en), eLife 9, doi: 10.7554/eLife.53898.
Week, Bob and Gideon Bradburd, Host-Parasite Coevolution in Continuous Space
Leads to Variation in Local Adaptation across Spatial Scales, The American
Naturalist 0, no. 0, 000–000, doi: 10 .1086/727470, eprint: https ://doi .org/
10.1086/727470.
Yoder, Jeremy B. and Scott L. Nuismer, When Does Coevolution Promote Diver-
sification? (en), The American Naturalist 176, no. 6, 802–817, doi: 10.1086/
657048.
46
CHAPTER 3
CHAPTER 3 THE GENETIC LANDSCAPES OF LAKE MALAWI CICHLIDS
The experimental question and design for this project was developed by my-
self, Dr. Anastasia Teterina, Aidan Short, and Dr. Peter L. Ralph. I gathered all
data, ran a majority of the analysis, and wrote the manuscript for this chapter
along with Dr. Peter L. Ralph. Dr. Anastasia Teterina advised on many steps of
data cleaning and analysis, and Aidan Short collaborated on producing some of the
introgression analysis. Both aided in the interpretation of results.
3.1 Introduction
Population genetics serves as a fundamental framework in the study of evolu-
tion, elucidating the mechanisms driving genetic variation within and between pop-
ulations (Lewontin 1974; Rodrigues, Kern, and Ralph 2024). Through the exami-
nation of genetic drift, gene flow, mutation, recombination, and natural selection,
population genetics has provided insights into how populations evolve and adapt
over time. These different forces have all played a role in shaping patterns of varia-
tion in genomes, and describing the “balance” between these forces is an important
emerging issue (Renaut et al. 2013).
Populations can be subject to different evolutionary forces at any point in their
lineages. These evolutionary forces can leave signals in the genetic landscapes, but
these signals are often combatant and easily distorted or complicated to measure
(Van Doren et al. 2017; Wang et al. 2020; Rodrigues, Kern, and Ralph 2024). Ad-
vances in sequencing technology have allowed examination of genetic landscapes of
closely related species, revealing peaks and valleys in genetic diversity and diver-
gence that were similar across species and species pairs. These peaks and valleys
were once thought to be markers of speciation (Cruickshank and Hahn 2014), but
can also occur due to the process of shared evolutionary history (Wang et al. 2020).
Patterns of genetic diversity and divergence have long been used to uncover
the molecular signs of evolution. For instance, comparisons between synonymous
and non-synonymous changes in coding sequence, either within species (the McDonald-
Kreitman test, McDonald and Kreitman 1991) or between species (dN/dS, Kimura
1977; Kryazhimskiy and Plotkin 2008), can indicate that evolution is either acting
to remove or encourage new mutations in a particular gene or set of genes. The
47
signal provided by these measures works primarily because natural selection di-
rectly affects the chance that each new mutation under selection rises in frequency
and goes to fixation: “positive selection” tends to increase frequencies of new al-
leles, while “negative selection” tends to decrease them. More indirectly, selection
often decreases genetic diversity in regions near to the loci at which alleles are un-
der direct selection (Kaplan, Hudson, and Langley 1989; Hudson and Kaplan 1995;
Charlesworth, Morgan, and Charlesworth 1993). This predicts a negative corre-
lation between genetic diversity and recombination rate that has been widely ob-
served (Langley et al. 2012; Corbett-Detig, Hartl, and Sackton 2015; Burri 2017).
However, these indirect (or, “linked”) effects of selection only affect levels of within-
species genetic diversity – only the direct effects of selection affect the rate at which
between-species fixations accumulate (Birky and Walsh 1988). In summary, positive
selection tends to decrease within-species diversity but increase the rate of accu-
mulation of between-species divergence as compared to neutral regions, while nega-
tive selection tends to decrease both. Furthermore, linked effects are expected to be
correlated with recombination rate, while direct effects are not. These differences
provide a possible avenue to understand the relative intensities of positive and neg-
ative selection on a genomic scale in a particular group of taxa (and for non-coding
regions), which is a long-standing open question in evolutionary biology.
However, the problem of deciphering signals is made more difficult because of a
complex array of other processes and patterns that can affect genetic diversity and
divergence differently along the genome. For instance, the direct effects of positive
selection increase the rate of between-species divergence, but since divergence also
depends on diversity within the ancestor, regions with a higher intensity of positive
selection may not actually have higher divergence. This makes it necessary to com-
pare divergences between increasingly distant taxa, rather than in only one or two
species. An increased mutation rate in one region of the genome can resemble the
net effects of direct selection, except that higher mutation rates will increase diver-
sity. GC-biased gene conversion also resembles the direct effects of positive selec-
tion (Duret and Galtier 2009), but its known dependence on particular nucleotide
signatures makes its effects identifiable (Rodrigues, Kern, and Ralph 2024) rather
than decreasing it. Balancing selection can strongly increase diversity, possibly over
large regions if recombination is suppressed, e.g., as for an inversion (Kirkpatrick
and Barton 2006; Kirkpatrick 2010).
Finally, much of this theory and intuition for genetic diversity and divergence
48
is based on models of a single species tree. However, even some well-separated species
interbreed, and although the picture of species as distinct, randomly-mating units
(possibly exchanging genetic material with each other) is a good description for
many purposes, it is not clear how well that description applies to some groups
of taxa. Furthermore, introgressing genetic material is subject to particular selec-
tive forces: for instance, gene flow may be more directional in gene-rich regions if
one species has a higher level of genetic load (Barton and Gale 1993; Bierne et al.
2002; Harrison and Larson 2016) or if one species has adapted to an environment
the other is moving into. Indeed, introgression has been shown in some well-studied
cases to be correlated with gene density and/or recombination rate (Harris and
Nielsen 2016).
Since there are a large number of confounding factors, to take full advantage
of the evolutionary information provided by genetic diversity and divergence, we
need to compare across regions of the genome, chromosomes, and species pairs of
increasing divergence. In particular, it is important to compare across different
groups of taxa that might have different confounders (e.g., gene density and recom-
bination rate might be positively or negatively correlated) and different balance of
forces (e.g., introgression might be more or less important). This will help us un-
derstand what signals are repeatable, how to control for various biological factors,
and how to extract a reliable signal.
Several studies of closely related taxa have found correlations along the genome
in genetic diversity (Burri et al. 2015; Irwin et al. 2016; Van Doren et al. 2017; Bat-
tey 2020). Studies have found such correlations not only between genetic diversity
within species but also between genetic divergences between species in Ficedula fly-
catchers (Burri et al. 2015), stonechats-flycatchers (Van Doren et al. 2017), great
apes (Rodrigues, Kern, and Ralph 2024) and Diplacus bush monkeyflowers (Stankowski
et al. 2019). These correlations of diversity and divergence persist over much longer
periods of time than what is expected under neutrality, implying that these corre-
lations are a reflection of shared processes acting in similar ways along the genomes
of these separate species.
The great apes studied by Rodrigues, Kern, and Ralph (2024) represents a
very well-studied system in which five species (human, chimpanzee, bonobo, go-
rilla and orangutan) are well-separated in evolutionary time: typical inter-node dis-
tances in the phylogeny are upwards of 10Ne, so there is very little shared ancestral
variation retained between the species. However, a number of subspecies are more
49
closely related, giving resolution at shorter time scales. Correlations in diversity or
divergence between closely related taxa were around 0.9, and decayed to around 0.4
over the depth of the phylogeny (about 12 million years). The conclusion (based
on comparison to simulations) was that correlations were likely maintained by a
combination of positive and negative selection, as well as a possible contribution of
mutation rate variation. Introgression is not thought to have a strong effect on the
genomes, and the genomes are generally very syntenic, which simplified the anal-
yses but did not provide the opportunity to study the effects of introgression or
rearrangements.
The group of Diplacus bush monkeyflowers studied by Stankowski et al. (2019)
is a relatively recent species radiation around 1Mya in (or near) southern Califor-
nia. Taxa show varying degrees of ability to interbreed, but many are geographi-
cally isolated, so the amount of ongoing hybridization is unclear. However, at least
some taxa have known hybrid zones (Sobel and Streisfeld 2015). In Diplacus mon-
keyflowers, gene density is negatively correlated with recombination rate, whereas
in great apes, this correlation is positive. Since the strength of linked selection in-
creases with gene density and decreases with recombination rate, this suggests that
the strength of linked selection varies much more across the genome in monkeyflow-
ers than it does in great apes. However, monkeyflower genomes are smaller than
those of great apes. Stankowski et al. (2019) found that the correlation between
FST and gene density strongly increases with the time separating the taxa, while
the correlation of FST and recombination rate strongly decreases. Furthermore, the
correlation between within-taxon genetic diversity and between taxon divergence
(dXY ) decreases from nearly 1 for closely related taxa to near zero for the most
distant taxa. Direct comparison between Stankowski et al. (2019) and Rodrigues,
Kern, and Ralph (2024) is difficult because they use different measures of genetic
diversity (e.g., FST versus dXY ) and of phylogenetic distance (time).
Both of the studies followed similar procedures with known phylogenetic rela-
tionships in different datasets highlighting unique aspects of genome evolution. To
add to this growing field, we examined the genetic landscapes of Lake Malawi hap-
lochromine cichlids. Similar to monkeyflowers, Lake Malawi cichlids are a recent
radiation with an estimated 500-860 species that diverged within the last 800,000
years (Brawand et al. 2014; Santos, Lopes, and Kratochwil 2023). The repeated
rapid expansion of cichlids is thought to be caused by cycles of expansion, diversi-
fication, and sexual selection (Kocher 2004; Wagner, Harmon, and Seehausen 2012;
50
Brawand et al. 2014). The genetic variation needed for such a rapid burst of speci-
ation might be due by frequent hybridization (Meier et al. 2017; Malinsky, Svardal,
et al. 2018; Svardal, Salzburger, and Malinsky 2021). This rapid expansion with a
large amounts of gene flow has made inferring genetic relationships difficult (Kocher
2004; Malinsky, Svardal, et al. 2018).
In addition to difficulties of reconstructing cichlid phylogenitic history, cichlids
are known to have large inversions (Conte et al. 2019; Svardal, Salzburger, and Ma-
linsky 2021), structural variation (Conte et al. 2019; Svardal, Salzburger, and Ma-
linsky 2021; Quah et al. 2024), and high rates of sex chromosome turnover (Conte
et al. 2019). Interestingly, cichlids are known to have low levels of diversity and
divergence even though they exhibit many phenotypic and behavioral differences
(Svardal, Salzburger, and Malinsky 2021). In this study we aim to investigate if ci-
chlids, a recent radiation, with large inversions and great amount of introgression,
show correlations in genetic diversity and divergence along the genome similar to
those seen in the great apes and Diplacus monkeyflowers. We also investigate how
genetic diversity and divergence interact with intrinsic properties of genomes (gene
density, recombination rate, site accessibility, gene repeats).
3.2 Methods
Samples and Alignment
We used publicly available data from Malinsky, Svardal, et al. (2018), by down-
loading sequencing data for each species with multiple individuals using SRA-toolkit
(fastq-dump.3.0.0), resulting in 53 individuals in 14 species (sample sizes in paren-
theses): Champsochromis caeruelus (2), Chilotilapia rhoadesii (4), Copadichromis
virginalis (7), Ctenopharynx intermedius (3), Dimidiochromis strigatus (2), Pro-
tomelas ornatus (4), Fossorochromis rostratus (4), Hemitilapia oxyrhynchus (2),
Lethrinops lethrinus (2), Mylochromis anaphyrmus (5), Otopharynx speciosus (2),
Placidochromis subocularis (7), Placidochromis cf longimanus (5), and Tremitochranus
placodon (4) (accession numbers in supplementary table B.1). Of these species, 13
are shallow benthic and one, C. virginalis, is utaka.
The 53 cichlid genomes were aligned to the Metriaclima zebra reference genome
(Ensembl Release 110, GCA 000238955.5) through the standard BWA-GATK pipeline
(https://gatk.broadinstitute.org). Reads were aligned using BWA-mem (v. 0.7.17-
51
r1188, Li and Durbin 2009) and unmapped reads were removed using samtools
(v. 1.6, Danecek et al. 2021). PCR dupicates were marked using picard v. 2.27.5-4-
g5295289-SNAPSHOT (http://broadinstitute.github.io/picard/). Coverage for each
individual was estimated using samtools depth. A list of indels were created us-
ing GATK v. 3.7 RealignerTargetCreator and then realigned using IndelRealigner.
Variants were called with GATK HaplotypeCaller v. 4.3.0.0 (McKenna et al. 2010).
Indels were marked by SelectVariants and then retained with VariantFiltration. Re-
peats were masked based on the available repeat mask for the genome from En-
sembl using generate_maked_ranges.py (https://gist.github.com/danielecook/
cfaa5c359d99bcad3200). High-quality indels and 10 base pairs around them, and
regions with bad coverage, specifically regions where the coverage was less than
three and more than three times the mean coverage for that sample in at least
75% of the samples were also masked. Chromosomes were divided into 1Mb non-
overlapping windows, and any windows that had less than 300Kb nonmissing data
were removed from the analysis using bedtools (Quinlan and Hall 2010).
For each 1Mb window we calculated accessibility, gene repeat density, gene
density, and recombination rate. “Accessible sites” are sites that are not masked
(i.e., no repeats, indels, or duplicates); we summarized “accessibility” for each win-
dow as the number of accessible sites divided by 1 million. Gene repeat density was
calculated using bedtools coverage with the file generated from generate_maked_ranges.py
and and the accessibility mask. Gene densities were also estimated with bedtools
coverage, a gene density file from the M. zebra Ensembl release (https://ftp.ensembl.
org/pub/release-110/gtf/maylandia_zebra_UMD2a.110.gff3), and again the ac-
cessibility mask. To obtain a genetic map, we obtained marker sequences from the
M. mbenjii × A. koningsi map from Conte et al. (2019), and used blast (v. 2.2.29,
Camacho et al. 2009) to find the coordinates of the markers on the M. zebra refer-
ence genome. We then calculated a rough estimation recombination rates for each
window (in cM/Mb) using genetic map distances from the genetic map of M. mben-
jii x A. koningsi from Conte et al. (2019). Two outlier recombination rates were
excluded from the study. All cM/Mb maps provided in appendix E.
Phylogeny
We first generated a phylip file from our filtered VCF file with vcf2phylip.py
(https://github.com/edgardomortiz/vcf2phylip) with default parameters, and then
52
generated a phylogenetic tree with all individuals using iqtree (v. 2.2.5 Nguyen et
al. 2015), using C. virginalis as the outgroup and 1000 bootstrap replicates. The
best fitting model selected by iqtree was TVM+F+R9, and the resulting phyloge-
netic tree grouped all individuals of the same species together (see Figure B.1).
To calculate a phylogenetic distance between species, we built a second tree
using only one individual per species in the same manner as above, using iqtree’s
best fitting model (TVM+F+I+R3). The individual with the best coverage was
selected for each species. For visualization purposes, we also estimated phylogenetic
trees separately by linkage groups.
Population genetics
We calculated genetic diversity within species (“π”) and genetic divergence be-
tween species (“dXY ”) in each 1Mb window using the windowed_divergence and
windowed_diversity functions from scikit-allel (v. 1.3.5, Miles et al. 2024). This
calculates both diversity and divergence as the proportion of accessible sites in the
window that differ between a pair of individuals, averaged either over pairs of in-
dividuals from the same species (for π) or over pairs of individuals with one from
each species (for dXY ).
We calculated Spearman’s correlations between genetic diversity landscapes,
divergence landscapes, and diversity and divergence landscapes in python. Spear-
man’s correlations between the genome functions (gene density, gene repeats, acces-
sibility, and recombination rate) and genetic divergence were also calculated.
We plotted phylogenetic trees using the ape package (v. 5.7-1, Paradis and
Schliep 2019) in R (v. 4.3.2, R Core Team 2021) and Rstudio (v. 2023.09.1+494).
Distance between tips were calculated using cophenetic and dispRity (v. 1.8, Guillerme
2018).
Figure 3.1 shows how phylogenetic distances between different pairs of diver-
sity and/or divergence are calculated in this study. The phylogenetic distance be-
tween species A and B, D(A,B), is used for π vs π correlations, and is shown in
Figure 3.1A:
D(A,B) = d(A,B), (3.1)
where A and B are populations and d is phylogenetic distance (shown in red in Fig-
ure3.1). The distance between a pair of species, A and B, with another species C,
D(AB,C), is used for dXY vs π correlations, and is shown in Figure 3.1B). D(AB,C)
53
Figure 3.1. An example phylogenetic tree with four populations (A, B, C, D). π is
calculated within one population, while dXY is calculated between two populations.
Yellow shading highlights the path used to calculate phylogenetic distance in equa-
tions (3.1), (3.2), and (3.3). In (A) the red path is the distance between population
A and B also seen in equation (3.1). In (B) the yellow path highlights the distance
between the node between A and B, and C. To calculate phylogenetic distance
we can use equation (3.2), which takes the path between A and C (purple) and B
and C (green) and takes out the distance between A and B (red). Since the two
paths overlap, the two is divided out. In (C) the yellow path highlights the dis-
tance between the node between A and B, the node between C and D. To calculate
phylogenetic distance we can use equation (3.3), which takes the path between A
and C (purple), B and C (green), A and D (blue), B and D (yellow) and takes out
the distance between A and B (red) and C and D (orange). Since the four paths
overlap, the four is divided out.
is the phylogenetic distance from species C to the path between A and B. This is
computed as:
d(A,C) + d(B,C)− d(A,B)
D(AB,C) = , (3.2)
2
where again A, B, and C are populations. This calculation adds the distance be-
tween A and C (purple), the distance between B and C (green), and then subtracts
the distance between A and B (red). Two paths overlap so a two is divided out.
The distance between two pairs of species, A–B and C–D, denoted D(AB,CD), is
used for dXY vs dXY correlations, and is shown in Figure 3.1C). It is the distance
in the tree between the path from A to B and the path from C to D, if these do not
overlap; if they do, it is minus one times the length of their overlap. It is computed
54
as follows:
d(A,C) + d(A,D) + d(B,C) + d(B,D)− 2d(A,B)− 2d(C,D)
D(AB,CD) = ,
4
(3.3)
This calculation adds the distance between A and C (purple), the distance between
A and D (blue), the distance between B and C (green), the distance between B and
D (yellow), then subtracts the distance between A and B (red) and between C and
D (orange) twice. There are four paths overlapping so a four is divided out.
Data visualisation and analyses were completed using ggtree (v. 10.3.0, Xu et
al. 2022) and ggplot2 (v. 3.4.4, Wickham 2009), phytools (v. 2.1-1, Revell 2024),
reshape2 (v. 1.4.4, Wickham 2007), and tidyr (v. 1.3.0, Wickham, Vaughan, and
Girlich 2023).
Introgression Analyses
To estimate the impact of introgression on genome-wide levels of genetic di-
vergence and diversity, we first tested for evidence of introgression by calculating
Patterson’s D (Green et al. 2010) for all possible trios of in-group taxa using the
Dtrios command in Dsuite (Malinsky, Matschiner, and Svardal 2021) based on the
relationships inferred from genomic data in our phylogenetic tree with C. virginalis
as the outgroup. To estimate the history of hybridization among these taxa we
then used Dsuite to identify recent and historical signatures of hybridization by cal-
culating the f -branch statistic based on all significant values of Patterson’s D (see
supplementary index I). To estimate genome-wide variation in introgression, we cal-
culated the admixture proportion (fd) (Martin, Davey, and Jiggins 2015) in 1Mb
non-overlapping windows for all possible trios of ingroup taxa that test for intro-
gression between a pair of taxa that display high levels of recent and historical in-
trogression based on f -branch values. To calculate fd we used the ABBABABAwin-
dows.py Python script (https://github.com/simonhmartin/genomics_general) with
three species (a “trio” P1, P2, and P3) and an outgroup (C. virginalis). fd measures
shared ancestry between P2 and P3 that is not shared by P1, and hence quantifies
introgression from P3 into P2 that has occurred since the divergence of P1 and P2.
We also calculated fd by linkage group
55
3.3 Results
Our study investigated the genetic landscape of cichlids through patterns of
similarities and differences in diversity and divergence in 1Mb windows along the
genome. Out of a total of 771 windows, 30 were excluded due to low levels of ac-
cessibility (see methods). We compared how the relationship between diversity and
divergence changed over a range of phylogenetic distances. Additionally, we exam-
ined the relationship between diversity and divergence with other components of
the genome (gene density, recombination, gene repeats, site accessibility), and how
these relate to measures of introgression.
Divergence across the genome
Divergence along all linkage groups in the genome is shown in Figure 3.2A.
Each point shows divergence between one pair of species over a 1Mb window (ge-
netic diversity for all linkage groups is shown in supplementary figure B.2). Diver-
gences between species were low but varied substantially along the genome, ranging
from about 0.1% to 0.4%, while diversity ranged from 0.05% to 0.3% similar to the
levels of diversity seen in Malinsky, Svardal, et al. (2018).
The divergences between different pairs of taxa along each linkage group were
highly correlated (plots of each individual linkage groups are shown in appendix C
and D). For instance, Figure 3.2B zooms into a typical linkage group, LG5, with
lines showing the divergence between each species pair. In this linkage group, as
in most others, divergence from all other taxa to C. virginalis is higher than diver-
gences between other pairs of taxa (comparisons to C. virginalis shown as green
lines). We use C. virginalis as the outgroup, although it does not show the highest
divergence to other taxa on all linkage groups. Divergence differs more between dif-
ferent windows along the genome than it does between different pairs of taxa, un-
like what was seen in great apes (Rodrigues, Kern, and Ralph 2024) or monkeyflow-
ers (Stankowski et al. 2019). However, the relative ordering of genetic diversity be-
tween different species pairs is highly consistent across windows (a feature also seen
in both great apes and monkeyflowers).
However, some linkage groups present different patterns than we saw on link-
age group 5. Figure 3.2C zooms into linkage group 11, where divergences between
each pair of species fall into one of three distinct groups. Furthermore (although
56
A Dxy by Window
0.006
0.004
0.002
0 200 400 600 800
Window
B LG 5: Dxy By Window C LG 11: Dxy By Window D LG 22: Dxy By Window
0.0030 0.006
0.0030
0.0025 0.0025
0.004
0.0020
0.0020
0.0015
0.0015 0.002
0.0010
150 160 170 370 380 390 700 710 720
Window Window Window
Figure 3.2. (A) Divergence in 1Mb windows between all species pairs on all linkage
groups, with divergence for each species pair in each window shown as a point. The
30 windows (out of a total of 771) omitted due to low accessibility are shown with
a grey background. (B,C,D) Divergences in 1Mb windows for linkage groups 5, 11,
and 22 respectively, with divergences for each species pair shown as a single line.
this is not evident in the plot), each individual species’ comparisons to other species
all fall either in the top and bottom, or in the top and middle groups. For instance,
comparisons to C. virginalis (which are all greenish lines) all fall in either the top
or middle group. This is consistent with the presence of an inversion, and indeed,
linkage group 11 is known to harbor a large inversion that is polymorphic between
species (Conte et al. 2019), and is associated with bower-building behavior (York et
al. 2018). Interestingly, divergence between the two orientations is similar to diver-
gence outside of the inversion: an alternative pattern would show that divergence
in an inversion is elevated due to long-term maintenance under balancing selection.
One interpretation of the decreased divergence within both orientations of the in-
version could be that the inversion was polymorphic in the ancestor or occurred
early in the radiation (as suggested by Conte et al. (2019)), and has since been pre-
vented from introgressing between taxa by sexual selection, unlike much of the rest
of the genome.
57
Dxy Dxy
Dxy
Dxy
Figure 3.2D zooms into linkage group 22, the linkage group with the window
having highest divergences (of those that were not removed). A window in this
linkage group shows much higher divergences between all pairs of taxa than do any
of the other windows; since the difference between this an other windows is so high,
it creates particularly high correlations between species comparisons on this linkage
group.
We do not see patterns as striking as linkage group 11 on any other linkage
group, even though Conte et al. (2019) reported large inversions on additional link-
age groups (2, 9 and 20). There is also the possibility that these patterns are dif-
ficult to perceive due to sex-determining loci (reported on linkage groups 7, 9 and
11), large region of low recombination (linkage groups 9), chromosomal fusion (link-
age group 7) (Conte et al. 2019).
Estimating phylogenetic distance
The whole-genome phylogeny estimated by iqtree is reasonably well-resolved,
and shown in Supplementary Figure B.3. Since it is thought that the relationships
between these cichlid species are not well-represented by a single phylogeny, we also
experimented with building local phylogenies in various ways. In particular, we also
estimated phylogenies separately using each linkage group; the resulting trees dif-
fered somewhat by linkage group, but were generally similar to Figure B.3 on all
linkage groups except LG11, where the inversion is visible (see appendix H). How-
ever, using linkage-group-specific phylogenies does not affect downstream results
substantially, so we proceeded with using the single, whole-genome phylogeny.
We do not claim that this phylogeny completely represents the history of these
cichlid species: at minimum, there are known to be a number of substantial intro-
gression events across the tree (so a phylogenetic network might be more appropri-
ate)(Santos, Lopes, and Kratochwil 2023). However, for downstream analyses it will
be useful to have a measure of “time” that separates different comparisons, and this
phylogeny should give us a good proxy for that notion.
Correlations between landscapes
Figure 3.3 shows all pairwise correlations between landscapes of diversity and
divergence, plotted against the distance between them in the phylogeny (calculated
58
A LG 5: Pi vs Pi Correlation B LG 5: Dxy vs Pi Correlation C LG 5: Dxy vs Dxy Correlation
1.0 1.0 1.0
0.9 0.9 0.9
0.8 0.8 0.8
0.7 0.7 0.7
0.6 0.6 0.6
0.5 0.5 0.5
0.03 0.04 0.05 0.06 0.07 0.02 0.03 0.04 0.004 0.008 0.012 0.016
   
D LG 11: Pi vs Pi Correlation E LG 11: Dxy vs Pi Correlation F LG 11: Dxy vs Dxy Correlation
1.00 1.00 1.00
0.75 0.75 0.75
0.50 0.50 0.50
0.25 0.25 0.25
0.00 0.00 0.00
0.03 0.04 0.05 0.06 0.07 0.02 0.03 0.04 0.004 0.008 0.012 0.016
Distance   
Figure 3.3. Correlations between different landscapes of genetic diversity or diver-
gence plotted against separating phylogenetic time (see text for details), for linkage
groups 5 and 11. (Left: π − π) Each point shows the correlation between land-
scapes of genetic diversity (π) for two taxa along linkage group 5, plotted against
the time separating the two taxa (i.e., the sum of the times from each of the two
taxa back to their MRCA). (Center: π − dXY ) Correlations between a landscape
of diversity in one taxon and divergence between two other taxa, plotted against
time from the first taxon to the path connecting the other taxa in the phylogenetic
tree. (Right: dXY − dXY ) Correlations between two landscapes of diversity, plotted
against separating phylogenetic distance (see text). Only comparisons with positive
separating phylogenetic distance are shown.
using equations (3.1), (3.2), or (3.3)) on linkage groups 5 and 11. For instance, Fig-
ure 3.2B shows the 14 landscapes of genetic diversity along linkage group 5, while
Figure 3.3A shows the 91 pairwise correlations between these 14 landscapes; the
x coordinate for the correlation between π(A) and π(B) is the distance between
species A and species B in the phylogeny. The x coordinate for correlations be-
tween diversity and divergence (Figure 3.3B) and between two landscapes of di-
vergence (Figure 3.3C) is found in a similar way as the distance in the phylogeny
between the paths connecting the species (see Methods).
Correlations are generally high, and then decrease with time. For instance,
correlations between landscapes of diversity on linkage group 5 all have a corre-
lation coefficient that starts around 0.95 and that drops off steadily with time.
This makes sense, as the effects of any process that produces or maintains sim-
ilarities between landscapes are expected to weaken with time: shared ancestral
59
Correlation Correlation
Correlation  
Correlation  
diversity will become less important as coalescence occurs in each taxon. Further-
more, we might expect selection to be acting with a very similar intensity on the
genomes of two recently diverged species, but selection will act more differently
across the genome the more the species diverge. We also see that correlations be-
tween (within-species) genetic diversity are much noisier than correlations between
(between-species) genetic divergence, a feature also noted in Rodrigues, Kern, and
Ralph (2024). This makes sense as local inbreeding and drift are expected to have a
larger effect on these patterns for diversity.
Although correlations decrease with time for both linkage groups, the correla-
tions with divergence for linkage group 11 form two distinct groups, each decreas-
ing. Furthermore, even some closely related comparisons on LG11 have correlations
near 0, something not seen on LG5. This is likely explained by the presence of an
inversion: as we saw in Figure 3.2, divergences between each pair of species fall
into one of three distinct groups, corresponding to whether the two species have
the same orientation of the inversion (and which orientation they have), or different
orientations. In other words, the two orientations of the inversion would correspond
to two long haplotypes, and we see the three possible comparisons between the two
haplotypes. We do not see a similar pattern for within-species diversity, likely due
to landscapes of diversity being strongly affected by genetic drift.
The example of linkage group 11 is useful because this gives us a view at a
large and well-resolved scale of the underlying process that acts mostly on a much
smaller genomic scale. For instance, if the inversion on LG11 was a much smaller
portion of the chromosome, the effect on correlations would be correspondingly
smaller: the two clouds in Figure 3.3E or F would be much closer together. How-
ever, if there are many instances of small regions of the genome (in inversions or
not) that move between species, the result could produce a more gradual decrease
of correlation with time (as in Figure 3.3B or C).
Although inversions have been reported on other linkage groups, we do not
see clearly separated groups like those shown for LG11 in Figure 3.2. However, we
do see patterns of banding similar to those of LG11 in Figure 3.3 for other linkage
groups, most strongly, LG2 and LG13 (see appendix F). It seems likely that the
same mechanism underlies these patterns.
Correlation between divergences of pairs of great ape species is only about 0.75
for the most closely related comparisons and decays to around 0.5 for the correla-
tions between the most distant comparisons (about 106 generations separated in
60
the tree). Correlations between cichlids, on the other hand are, excluding linkage
groups with likely inversions, mostly much higher: above 0.8 even for the most dis-
tant comparisons (which are separated by a few hundreds of thousands of genera-
tions). One reason for this is likely the relevant time scale: the largest phylogenetic
distances between cichlids is more comparable to the smaller phylogenetic distances
between great apes. However, it is likely that ongoing introgression between cichlids
also plays a strong role.
Correlations with genomic features
Genetic diversity and divergence are expected to be correlated with genomic
features such as gene density and recombination, because these features regulate
the strength of linked selection. Indeed, as shown in Figure 3.4, the genomic fea-
tures we studied (gene density, recombination rate, accessibility, and repeat con-
tent) are often correlated with landscapes of diversity and divergence: on many
linkage groups, correlations are in the range 0.2–0.5. There were significant correla-
tions between averaged windowed divergence and gene repeats (p-value 3.05× 10−7,
correlation 0.188522), averaged windowed divergence and accessibility (p-value
< 2.2 × 10−16, correlation -0.4534826), and averaged windowed divergence and re-
combination rate (p-value 0.0122, correlation 0.09879411). However, the value and
even sign of those correlations differs substantially by linkage group.
Appendix G shows these correlations against time: in other words, if dXY (A,B)i
is the divergence between species A and B in window i and fi is a genomic feature
of that window, then cor(dXY (A,B), f) is plotted against the phylogenetic time
separating A and B. Stankowski et al. (2019) and Rodrigues, Kern, and Ralph
(2024) found striking trends with time in such plots, but our results show no con-
sistent trends with time.
Although correlations do not change with phylogenetic distance, they do differ
substantially between linkage groups. This is quite different to what Stankowski et
al. (2019) saw: correlations were similar across all Diplacus chromosomes. However,
correlations with recombination rate and exon density also varied substantially by
chromosome in the great apes (data replotted from Rodrigues, Kern, and Ralph
(2024) in Supplementary Figure B.5). This emphasizes the need for more examples
of this sort of data: even within cichlids, if we had restricted attention to linkage
group 1 (which has a correlation near 0.4 with recombination rate) we might con-
61
Correlation Plots by Linkage Group
Dxy by Gene Density
0.4
0.0
−0.4
−0.8
Dxy by Gene Repeats
0.4
0.0
−0.4
−0.8
Dxy by Accesibility
0.4
0.0
−0.4
−0.8
Dxy by Recombination
0.4
0.0
−0.4
−0.8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 22 23
LG
Figure 3.4. Shows the correlation between four genome functions (gene density,
gene repeats, accessibility, recombination) and dXY for each linkage group. In each
boxplot, correlations between the genome functions and dXY contains information
from all species pairs. The correlation between gene density and dXY range from
-0.5 to 0.5, with about half of the correlations crossing through 0. The correlation
between gene repeats and dXY range from -0.25 to 0.75, with a majority of the cor-
relations being positive. The correlation between accessibility and dXY range from
-0.75 to 0.25, with a majority of the correlations being negative. The correlation
between recombination and dXY range from -0.5 to 0.5, with a majority of the cor-
relations being close to 0. There are bimodal distributions of correlation values in
several linkage groups with the most noticeable occurring in linkage groups 2, 11,
and 13.
62
Correlation
clude something very different than for linkage group 8 (which has a correlation
around -0.4 with recombination rate). For cichlids, it seems that large-scale features
are much more stochastic than in other species.
Supplementary Figure B.4 shows correlations of divergence with genomic fea-
tures, separated not by linkage group but by species comparison.
Introgression
Consistent with previous studies on the history of hybridization and introgres-
sion among Lake Malawi cichlids, the f -branch statistic showed evidence of ex-
tensive recent and historical introgression among the taxa of this radiation. The
strongest signals are of recent hybridization between speciosus and caeruelus, as
well as historical hybridization between speciosus and the common ancestor of caeru-
elus, strigatus, rhoadesii, ornatus, oxyrhynchus, subocularis, and placodon (observed
in heat maps seen in appendix I).
To investigate how these histories affect landscapes of diversity and divergence,
we used fd to measure the degree of shared ancestry (i.e., the amount of historical
introgression) in windows along the genome among various species pairs. Calcu-
lations of fd performed using taxa at increasing levels of divergence should iden-
tify introgression that occurred further back in time, because fd can only identify
introgression that has occurred since the divergence of the taxa used as P1 and
P2 for the calculation of fd (Martin, Dasmahapatra, et al. 2013; Short and Streis-
feld 2023). Roughly speaking, fd measures amount of ancestry that is shared by
two taxa, P2 and P3, that is not shared by another taxon, P1; taking P1 and P2 to
be sister taxa in the phylogeny, this indicates historical gene flow from P3 to P2
more recently than the divergence between P1 and P2. Figure 3.5 shows fd from
speciosus (P3) to caeruelus (P2) since the divergence of strigatus, rhoadesii, orna-
tus, oxyrhynchus, subocularis, placodon, or rostratus (P1).
Figure 3.5A shows evidence of higher levels of historical introgression when the
most genetic diverged taxon, rostratus, was used as P1. We ran additional fd cal-
culations to identify introgression that has occurred since the divergence of specio-
sus and its sister taxon, intermedius (which display similar relative divergences to
rostratus and caeruelus, see Supplementary Figure B.3) and also identified simi-
larly high levels of introgression. Our identification of greater fd values in these
two trios containing caeruelus, suggests that hybridization has occurred at multi-
63
A Whole genome mean fd
Trio
P1: strigatus P2: caeruelus P3: speciosus
0.04
P1: rhoadesii P2: caeruelus P3: speciosus
P1: ornatus P2: caeruelus P3: speciosus
P1: oxyrhynchus P2: caeruelus P3: speciosus
0.02
P1: subocularis P2: caeruelus P3: speciosus
P1: placodon P2: caeruelus P3: speciosus
P1: rostratus P2: caeruelus P3: speciosus
0.00
0.00164 0.00166 0.00168 0.00170 0.00172 0.00174
mean dxy between P1 and P2
B LG 2: fd by window C LG 19: fd by window
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 10000000 20000000 30000000 0 5000000 10000000 15000000 20000000 25000000
Window Window
Figure 3.5. (A) Average fd for 1Mb windows across the entire genome, colored
by the taxa used for P1. (B,C) fd for the same sets of taxa in 1Mb windows for
linkage groups 2 and 19 respectively, with fd for each trio shown as a single line.
ple time points throughout their divergence history.
Figure 3.5B and C shows that the signal of introgression varied by linkage
group. For example, fd is larger on linkage group 2 than on linkage group 19. Fur-
thermore, consistent with our genome-wide estimates of fd, we identified greater
values on linkage group 2 using rostratus as P1 (fd > 0.0625) as compared to taxa
that are less diverged from caeruelus (< 0.05). On Chromosome 19 however we
identified no such pattern with the average fd value consistently remaining between
0.02 and 0.04 regardless of which taxon was used as P1.
We identified a negative correlation between fd and dXY between rostratus and
caeruelus (≈ −0.103), and a strong, negative correlation between fd and dXY be-
tween caeruelus and speciosus (≈-0.320) (Supplementary figure B.6), which is con-
sistent with introgression occurring from speciosus into caeruelus. This suggests
that introgression is largely acting to decrease divergence between caeruelus and
speciosus as well as between rostratus and caeruelus. Surprisingly, we identified a
64
fd mean f d
fd
strong negative correlation between fd and π for both speciosus and caeruelus (≈-
0.2), which raises the possibility that widespread adaptive introgression has con-
tributed to the retention of introgressed alleles across the genome of these species.
3.4 Discussion
In this study, we examined genetic diversity and divergence in 1Mb windows
along the genomes of 14 cichlid species. Genetic diversity was low but varied be-
tween and within linkage groups, ranging from around 0.05% to 0.3%, and diver-
gence ranged from 0.1% to 0.4%, similar to values seen in Malinsky, Svardal, et al.
(2018) and Svardal, Salzburger, and Malinsky (2021). Both within-species diver-
sity and between-species divergence had very similar valleys and peaks across most
linkage groups, although a striking pattern of three distinct “pathways” is visible on
LG11, likely due to a polymorphic inversion reported in Conte et al. (2019).
Although correlations between landscapes of diversity or divergence were high
– mostly above 0.7 – correlations decreased with time, i.e., were lower between
more phylogenetically distant comparisons. The patterns and levels of correlation
differed by linkage group, as for example on LG11. This decrease of correlation
with time has also been observed in other groups, e.g., great apes (Rodrigues, Kern,
and Ralph 2024) and bush monkeyflowers (Stankowski et al. 2019).
To understand the balance of forces that maintain these correlations between
diversity and divergence it can help to examine the relationship to other genomic
features. Across the cichlid genomes we examined, genetic diversity and divergence
were correlated with recombination rate, repeat content, and accessibility, although
the strength of the correlations differed between linkage groups. Furthermore, the
correlation of any of these features with divergence did not strongly increase or
decrease with phylogenetic distance between the taxa, unlike in monkeyflowers
(Stankowski et al. 2019) or great apes (Rodrigues, Kern, and Ralph 2024).
We observed that levels of introgression differed between trios and varied be-
tween linkage groups. The patterns that are present in this dataset are likely caused
by complex effects of inversions, introgression, sexual selection, rearrangements and
linked selection discussed in further detail below.
65
Features of genomic landscapes
As reviewed above, one of the most commonly observed features of genomic
landscapes is a positive correlation between genetic diversity and recombination
rate (Corbett-Detig, Hartl, and Sackton 2015). This is taken as strong evidence for
the effects of linked selection, which also predicts a negative correlation of genetic
diversity and gene density. Indeed, genetic diversity is positively correlated with
recombination rate and negatively correlated with gene density genome-wide both
in our study as well as in Burri et al. (2015), Stankowski et al. (2019), Rodrigues,
Kern, and Ralph (2024), and Wang et al. (2020). The same is true of genetic diver-
gence as well in most of these studies. However, the correlations we see differ sub-
stantially by linkage group: as shown in Figure 3.4, these correlations when looked
at on different linkage groups can be of opposite sign, and sometimes quite large.
It is not usually reported whether these correlations differ so much by chromosome
in other taxa: at least in great apes, the data from Rodrigues, Kern, and Ralph
(2024) shows substantial variation between chromosomes, but a consistent sign.
This fact, and the fact that these correlations tend to be relatively weak (usually
less than 0.4) suggests that the intensity of linked selection is likely changing a lot
over time (e.g., as species radiate and genomes rearrange).
Chromosomal rearrangements
A number of large chromosomal rearrangements such as inversions and cen-
tromere repositioning events have been reported in cichlid species, as well as smaller-
scale gene duplications (Conte et al. 2019; Svardal, Salzburger, and Malinsky 2021).
The prevalence of these large rearrangements seems to be higher than in the other
taxa we compare to: for instance, there is only a single chromosomal fusion event
separating humans from other great apes and few other rearrangements, and chro-
mosomal structure of Fidecula flycatchers is very well-conserved (Burri et al. 2015).
The degree of chromosomal synteny in Diplacus (Stankowski et al. 2019) and Pop-
ulus (Wang et al. 2020) is less clear, but at least fewer instances of inversions or
other large rearrangements have been reported. Speculatively, rearrangements might
be having a stronger effect on cichlids than on these other taxa.
Perhaps the most obvious signal produced by a rearrangement would be that
of a relatively old inversion: if the inversion is older than typical coalescence time,
66
the two orientations of the inversion generally manifest as two distinct haplotypes,
so that genetic variation tends to nest within the two distinct groups defined by the
orientations. This agrees with what we see on linkage group 11. However, a more
recent inversion would look different: then, the inverted haplotype would be nested
within genetic variation on non-inverted haplotypes, and so have less obvious ef-
fects. Furthermore, smaller inversions produce smaller signals. It is interesting
to see how a large inversion creates discontinuous patterns in the relationship be-
tween time and correlation (e.g., Figure 3.3); the net effects of many smaller events
would create a gradual decrease. Since the putative inversion on linkage group 11
seems to be relatively old and show a pattern that does not agree with the genome-
wide phylogeny, this may suggest that it pre-dated the radiation (as suggested by
Svardal, Salzburger, and Malinsky (2021)). Many other rearrangements are known
to be present in cichlids, but we may not expect to see their effects if they are not
polymorphic in the taxa we study.
Sex determining regions
Cichlids are known to have an array of different sex determining loci and sys-
tems, with a high rate of turnover (Santos, Lopes, and Kratochwil 2023). In Lake
Malawi cichlids alone, five sex determining systems have been noted and described
(Feller et al. 2021). Both ZW and XY sex determining systems have been found
within the same species of Metriaclima (Ser, Roberts, and Kocher 2010). Such
rapid change in sex determination could create very different patterns of gene flow
and genetic diversity across chromosomes and across time. Incompatibilities be-
tween different sex determination systems can be coupled with a reduction in or
complete suppression of recombination (Conte et al. 2019). Changing the sex deter-
mination system can also skew the sex ratio (Feller et al. 2021), which impacts pop-
ulation dynamics. Selection pressures acting on sex-linked traits may drive adap-
tation and divergence, leading to the emergence of novel phenotypes (e.g., cichlid
color patterns) and speciation events (Wagner, Harmon, and Seehausen 2012; Conte
et al. 2019). Changes in sex determination across different cichlid species might
have contributed to the rapid radiation of cichlids (Feller et al. 2021) and could be
a major reason that we see such strikingly different patterns in genomic landscapes
and their rates of change across linkage groups.
67
Hybridization and introgression
It is well known that cichlids can hybridize with one another and that this
tendency to hybridize is thought to be partially responsible for the success of sev-
eral cichlid radiations (Meier et al. 2017; Malinsky, Svardal, et al. 2018; Svardal,
Salzburger, and Malinsky 2021). Repeated hybridization between both near and
distant lineages has led to introgression in between cichlid lineages, leading to com-
plex and difficult-to-infer historical relationships (Santos, Lopes, and Kratochwil
2023). Cichlid species with a history of introgression tend to have higher genetic di-
versity in some species groups, but most genetic variation present in the ancestral
Lake Malawi cichlids has been lost (Svardal, Salzburger, and Malinsky 2021). We
found signals of introgression between several species of cichlid, that differed be-
tween linkage groups (Figure 3.5). In particular, there is one large several megabase
signal on linkage group 2 (visible in two species comparisons), highlighting that in-
trogression can be relatively even or effect large regions. Interestingly, introgression
is usually expected to increase genetic diversity, but we observed less genetic diver-
sity in regions of the genome with more evidence of introgression (Supplementary
Figure B.6). This observation could be due to positive selection on introgressed al-
leles, or potentially other factors. Finally, the concept of introgression depends on
the existence of a species tree, and so a different framework might better describe
the recent dynamics of these species.
Limitations and continuing questions
We found that genomic landscapes of diversity and divergence were more sim-
ilar between species that were more closely related according to the genome-wide
phylogeny of Figure B.3. However, there is not a single phylogenetic relationship
that encompasses the entire genome: different portions of the genome have some-
times very different phylogenies. By using the phylogeny estimated from the whole
genome, we hope to get a reasonable description of relationships: e.g., which taxa
tend to be more closely related on most of the genome. In particular, the choice of
outgroup is well-supported across the genome, since it has noticeably higher diver-
gence to all other species on most linkage groups (see appendix C). A somewhat
different set of species or a different phylogenetic inference method would lead to a
different phylogeny, but we do not expect the main observations to be affected. It
68
could be better to analyze the data without reference to a phylogeny, but it seems
difficult to describe relationships between species otherwise.
There are also certainly some degree of errors generated by aligning to a sin-
gle reference genome. Some of the potential problems caused through the alignment
process can create false regions of diversity or divergence, e.g., a duplication that is
not present in the reference genome, or unmasked repetitive regions. Perhaps more
concerningly, hyper-divergent portions of the genome might not map well, thus re-
ducing divergence in the region. Furthermore, we simply ignore any non-syntenic
regions of the genome: it would be interesting to see how the amount of such re-
gions increases with time. In particular, it would be interesting to survey changes
in repetitive element content or location, along the lines of Quah et al. (2024) or
Stitzer et al. (2021).
Cichlids have attracted a lot of attention from evolutionary biologists due to
the phenotypic diversity generated by their recent adaptive radiation. The cich-
lid genome reflects this complex evolutionary history, with striking patterns that
appear at chromosomal scale along the genome. Patterns that we see are different
from what is observed in other species, so it is interesting to speculate that these
differences are related to the history and ecology of cichlids. For instance: is their
propensity for speciation reflected in patterns of genetic variation over time? Would
we see similar patterns in other taxa that have recently radiated (e.g., tomatoes,
pupfish, etc)? Further research integrating genomic data with ecological and geo-
logical information will help elucidate the links between cichlid evolution, genetic
variation, and speciation dynamics. Simulations could also be used to determine
how these different evolutionary forces interplay with one another to leave various
patterns in genetic data. Cichlids present a challenging case study about the rela-
tionship between evolutionary dynamics and genetic variation. This study thus pro-
vides the field with an additional set of observations for a species with a complex
evolutionary history.
3.5 References
Barton, Nicholas H and Katherine S Gale, Genetic analysis of hybrid zones, Hybrid
zones and the evolutionary process, Oxford University Press, 13–45.
69
Battey, Christopher J., Evidence of linked selection on the Z chromosome of hy-
bridizing hummingbirds* (en), Evolution 74, no. 4, 725–739, doi: 10.1111/evo.
13888.
Bierne, N et al., Deleterious mutations in a hybrid zone: can mutational load de-
crease the barrier to gene flow?, Genet Res 80, no. 3, 197–204.
Birky, C W and J B Walsh, Effects of linkage on rates of molecular evolution. (en),
Proceedings of the National Academy of Sciences 85, no. 17, 6414–6418, doi:
10.1073/pnas.85.17.6414.
Brawand, David et al., The genomic substrate for adaptive radiation in African ci-
chlid fish (en), Nature 513, no. 7518, 375–381, doi: 10.1038/nature13726.
Burri, R., Dissecting differentiation landscapes: a linked selection’s perspective, Jour-
nal of Evolutionary Biology 30, no. 8, 1501–1505, doi: 10 . 1111 / jeb . 13108,
eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jeb.13108.
Burri, Reto et al., Linked selection and recombination rate variation drive the evo-
lution of the genomic landscape of differentiation across the speciation contin-
uum of Ficedula flycatchers (en), Genome Research 25, no. 11, 1656–1665, doi:
10.1101/gr.196485.115.
Camacho, Christiam et al., BLAST+: architecture and applications (en), BMC
Bioinformatics 10, no. 1, 421, doi: 10.1186/1471-2105-10-421.
Charlesworth, B, M T Morgan, and D Charlesworth, The effect of deleterious mu-
tations on neutral molecular variation. (en), Genetics 134, no. 4, 1289–1303,
doi: 10.1093/genetics/134.4.1289.
Conte, Matthew A et al., Chromosome-scale assemblies reveal the structural evo-
lution of African cichlid genomes (en), GigaScience 8, no. 4, doi: 10 . 1093 /
gigascience/giz030.
Corbett-Detig, Russell B., Daniel L. Hartl, and Timothy B. Sackton, Natural Selec-
tion Constrains Neutral Diversity across A Wide Range of Species (en), PLOS
Biology 13, no. 4, e1002112, doi: 10.1371/journal.pbio.1002112.
Cruickshank, Tami E. and Matthew W. Hahn, Reanalysis suggests that genomic
islands of speciation are due to reduced diversity, not reduced gene flow (en),
Molecular Ecology 23, no. 13, 3133–3157, doi: 10.1111/mec.12796.
70
Danecek, Petr et al., Twelve years of SAMtools and BCFtools (en), GigaScience 10,
no. 2, giab008, doi: 10.1093/gigascience/giab008.
Duret, Laurent and Nicolas Galtier, Biased Gene Conversion and the Evolution of
Mammalian Genomic Landscapes, Annual Review of Genomics and Human Ge-
netics 10, no. 1, 285–311, doi: 10.1146/annurev-genom-082908-150001, eprint:
http://www.annualreviews.org/doi/pdf/10.1146/annurev-genom-082908-150001,
PMID: 19630562.
Feller, Anna F. et al., Identification of a novel sex determining chromosome in cich-
lid fishes that acts as XY or ZW in different lineages (en), Hydrobiologia 848,
no. 16, 3727–3745, doi: 10.1007/s10750-021-04560-7.
Green, Richard E et al., A draft sequence of the Neandertal genome, science 328,
no. 5979, 710–722.
Guillerme, Thomas, dispRity: A modular R package for measuring disparity (en),
Methods in Ecology and Evolution 9, no. 7, 1755–1763, doi: 10 . 1111/2041 -
210X.13022.
Harris, Kelley and Rasmus Nielsen, The Genetic Cost of Neanderthal Introgression,
Genetics 203, no. 2, 881–891.
Harrison, R G and E L Larson, Heterogeneous genome divergence, differential in-
trogression, and the origin and structure of hybrid zones, Mol Ecol 25, no. 11,
2454–2466, doi: 10.1111/mec.13582.
Hudson, R. R. and N. L. Kaplan, Deleterious background selection with recombina-
tion, Genetics 141, no. 4, 1605–1617.
Irwin, Darren E. et al., Recurrent selection explains parallel evolution of genomic
regions of high relative but low absolute differentiation in a ring species (en),
Molecular Ecology 25, no. 18, 4488–4507, doi: 10.1111/mec.13792.
Kaplan, N L, R R Hudson, and C H Langley, The "hitchhiking effect" revisited.
(en), Genetics 123, no. 4, 887–899, doi: 10.1093/genetics/123.4.887.
Kimura, M, Preponderance of synonymous changes as evidence for the neutral the-
ory of molecular evolution, Nature 267, no. 5608, 275–276, doi: 10.1038/267275a0.
Kirkpatrick, M, How and why chromosome inversions evolve, PLoS Biol 8, no. 9,
doi: 10.1371/journal.pbio.1000501.
71
Kirkpatrick, Mark and Nick Barton, Chromosome Inversions, Local Adaptation and
Speciation, Genetics 173, no. 1, 419–434, doi: 10 .1534/genetics .105 .047985,
eprint: https://academic.oup.com/genetics/article-pdf/173/1/419/42222075/
genetics0419.pdf.
Kocher, Thomas D., Adaptive evolution and explosive speciation: the cichlid fish
model (en), Nature Reviews Genetics 5, no. 4, 288–298, doi: 10.1038/nrg1316.
Kryazhimskiy, Sergey and Joshua B Plotkin, The population genetics of dN/dS,
PLoS genetics 4, no. 12, e1000304–e1000304, doi: 10 . 1371 / journal . pgen .
1000304.
Langley, C H et al., Genomic variation in natural populations of Drosophila melanogaster,
Genetics 192, no. 2, 533–598, doi: 10.1534/genetics.112.142018.
Lewontin, R.C., The Genetic Basis of Evolutionary Change, Biological Series, Columbia
University Press.
Li, Heng and Richard Durbin, Fast and accurate short read alignment with Bur-
rows–Wheeler transform (en), Bioinformatics 25, no. 14, 1754–1760, doi: 10 .
1093/bioinformatics/btp324.
Malinsky, Milan, Michael Matschiner, and Hannes Svardal, Dsuite-Fast D-statistics
and related admixture evidence from VCF files, Molecular ecology resources 21,
no. 2, 584–595.
Malinsky, Milan, Hannes Svardal, et al., Whole-genome sequences of Malawi cich-
lids reveal multiple radiations interconnected by gene flow (en), Nature Ecology
& Evolution 2, no. 12, 1940–1955, doi: 10.1038/s41559-018-0717-x.
Martin, Simon H, Kanchon K Dasmahapatra, et al., Genome-wide evidence for spe-
ciation with gene flow in Heliconius butterflies, Genome research 23, no. 11,
1817–1828.
Martin, Simon H, John W Davey, and Chris D Jiggins, Evaluating the use of ABBA–
BABA statistics to locate introgressed loci, Molecular biology and evolution 32,
no. 1, 244–257.
McDonald, John H. and Martin Kreitman, Adaptive protein evolution at the Adh
locus in Drosophila (en), Nature 351, no. 6328, 652–654, doi: 10.1038/351652a0.
72
McKenna, Aaron et al., The Genome Analysis Toolkit: A MapReduce framework
for analyzing next-generation DNA sequencing data (en), Genome Research 20,
no. 9, 1297–1303, doi: 10.1101/gr.107524.110.
Meier, Joana I. et al., Ancient hybridization fuels rapid cichlid fish adaptive radia-
tions (en), Nature Communications 8, no. 1, 14363, doi: 10.1038/ncomms14363.
Miles, Alistair et al., cggh/scikit-allel: v1.3.8, doi: https ://zenodo.org/records/
10876220.
Nguyen, Lam-Tung et al., IQ-TREE: A Fast and Effective Stochastic Algorithm
for Estimating Maximum-Likelihood Phylogenies (en), Molecular Biology and
Evolution 32, no. 1, 268–274, doi: 10.1093/molbev/msu300.
Paradis, Emmanuel and Klaus Schliep, ape 5.0: an environment for modern phylo-
genetics and evolutionary analyses in R (en), Bioinformatics 35, no. 3, 526–528,
doi: 10.1093/bioinformatics/bty633.
Quah, Fu Xiang et al., A pangenomic perspective of the Lake Malawi cichlid radia-
tion reveals extensive structural variation driven by transposable elements (en),
doi: 10.1101/2024.03.28.587230.
Quinlan, Aaron R. and Ira M. Hall, BEDTools: a flexible suite of utilities for com-
paring genomic features (en), Bioinformatics 26, no. 6, 841–842, doi: 10.1093/
bioinformatics/btq033.
R Core Team, R: A Language and Environment for Statistical Computing, R Foun-
dation for Statistical Computing, Vienna, Austria.
Renaut, S. et al., Genomic islands of divergence are not affected by geography of
speciation in sunflowers (en), Nature Communications 4, no. 1, 1827, doi: 10.
1038/ncomms2833.
Revell, Liam J., phytools 2.0: an updated R ecosystem for phylogenetic comparative
methods (and other things) (en), PeerJ 12, e16505, doi: 10.7717/peerj.16505.
Rodrigues, Murillo F, Andrew D Kern, and Peter L Ralph, Shared evolutionary
processes shape landscapes of genomic variation in the great apes (en), Genet-
ics 226(4), doi: 10.1093/genetics/iyae006.
Santos, M. Emília, João F. Lopes, and Claudius F. Kratochwil, East African cichlid
fishes (en), EvoDevo 14, no. 1, 1, doi: 10.1186/s13227-022-00205-5.
73
Ser, Jennifer R., Reade B. Roberts, and Thomas D. Kocher, MULTIPLE INTER-
ACTING LOCI CONTROL SEX DETERMINATION IN LAKE MALAWI CI-
CHLID FISH (en), Evolution 64, no. 2, 486–501, doi: 10.1111/j.1558- 5646.
2009.00871.x.
Short, Aidan W and Matthew A Streisfeld, Ancient hybridization leads to the re-
peated evolution of red flowers across a monkeyflower radiation, Evolution Let-
ters 7, no. 5, 293–304.
Sobel, James M. and Matthew A. Streisfeld, Strong premating reproductive iso-
lation drives incipient speciation in Mimulus aurantiacus: REPRODUCTIVE
ISOLATION IN M. AURANTIACUS (en), Evolution 69, no. 2, 447–461, doi:
10.1111/evo.12589.
Stankowski, Sean et al., Widespread selection and gene flow shape the genomic
landscape during a radiation of monkeyflowers (en), PLOS Biology 17, no. 7,
e3000391, doi: 10.1371/journal.pbio.3000391.
Stitzer, Michelle C. et al., The genomic ecosystem of transposable elements in maize,
PLOS Genetics 17, no. 10, 1–30, doi: 10.1371/journal.pgen.1009768.
Svardal, Hannes, Walter Salzburger, and Milan Malinsky, Genetic Variation and
Hybridization in Evolutionary Radiations of Cichlid Fishes (en), Annual Review
of Animal Biosciences 9, no. 1, 55–79, doi: 10.1146/annurev-animal-061220-
023129.
Van Doren, Benjamin M. et al., Correlated patterns of genetic diversity and differ-
entiation across an avian family (en), Molecular Ecology 26, no. 15, 3982–3997,
doi: 10.1111/mec.14083.
Wagner, Catherine E., Luke J. Harmon, and Ole Seehausen, Ecological opportu-
nity and sexual selection together predict adaptive radiation (en), Nature 487,
no. 7407, 366–369, doi: 10.1038/nature11144.
Wang, Jing et al., Evidence for widespread selection in shaping the genomic land-
scape during speciation of Populus (en), Molecular Ecology 29, no. 6, 1120–
1136, doi: 10.1111/mec.15388.
Wickham, Hadley, ggplot2: Elegant Graphics for Data Analysis (en), Springer New
York, New York, NY, doi: 10.1007/978-0-387-98141-3.
, Reshaping Data with the reshape Package (en), Journal of Statistical Soft-
ware 21, no. 12, doi: 10.18637/jss.v021.i12.
74
Wickham, Hadley, Davis Vaughan, and Maximilian Girlich, tidyr: Tidy Messy Data,
R package version 1.3.0.
Xu, Shuangbin et al., Ggtree : A serialized data object for visualization of a phylo-
genetic tree and annotation data (en), iMeta 1, no. 4, e56, doi: 10.1002/imt2.
56.
York, Ryan A. et al., Behavior-dependent cis regulation reveals genes and pathways
associated with bower building in cichlid fishes (en), Proceedings of the National
Academy of Sciences 115, no. 47, doi: 10.1073/pnas.1810140115.
75
CHAPTER 4
CONCLUSION
This dissertation describes my work investigating the many different signals
produced by the process of evolution. The evolutionary mechanisms of natural se-
lection, adaptation, mutation, gene flow, genetic drift, and non-random mating all
interact in various ways over multiple timescales. These mechanisms have left both
ecological and molecular patterns in all life on Earth. The goal of this dissertation
was to observe and examine these pattens to expand our current knowledge on how
the processes of evolution impact populations.
In my second chapter, I investigated the evolutionary trajectories of a coevo-
lutionary arms race between a predator (snakes) and its’ prey (newts). I used spa-
tial simulations to test the effects of genetic basis and spatial heterogeneity on trait
evolution. I found that newts and snakes could coevolve under most genetic archi-
tectures, except when the mutational variance of a specific architecture was too
small. Mutational variance was an important factor that greatly impacted coevo-
lutionary trajectories. These results agree with previous research confirming how a
lack of genetic variation created by low mutation variance can effectively halt evo-
lution is some coevolutionary interactions. As I explored the geographic variation
of newts’ toxicity and snakes’ resistance across space and time, I found that my
simulations could only produce coevolutionary mosaics if there was spatial hetero-
geneity existing within the simulation regardless of the genetic architecture. The
result from these simulations highlights the importance of including spatial aspects
into the study of evolution, especially when dealing with larger geographical areas.
Findings from this chapter impacts the field of coevolution theory by improving our
understanding of how the genetic basis of traits can impact coevolutionary trajecto-
ries and by showing the importance of including geographical space when studying
evolution.
In my third chapter, I switch from exploring evolutionary processes though
simulation to observing the signals of past evolution by examining the landscapes
of cichlid genomes. The many different mechanisms of evolution have left patterns
that can be interpreted. In cichlids I found that there were low levels of genetic di-
versity and divergence across the genome. Even though the levels of genetic diver-
sity and divergence were low, I observed that they were strongly correlated. I found
that as species became more phylogenetically different these correlations weakened.
76
There were also strong correlations between levels of genetic divergence and genome
functions. The most interesting finding from this study was the striking differences
between linkage groups. Some of the locations of these striking patterns align with
previous research and point to these signals coming from a combination of inver-
sions and sexual selection. Although these signals provide clues about the history
of evolutionary forces acting on these genomes, this history is sufficiently complex
that we cannot yet see the entire picture from these data. The findings from my
second chapter fit well within our current knowledge of genetic landscapes espe-
cially when compared to the genomic landscapes of great apes and monkeyflowers.
The research that I did for this chapter adds additional knowledge point into this
growing field.
My dissertation encompasses a small, but specific portion of the field of evolu-
tion. The findings in my dissertation contributed to our understanding of evolution-
ary signals of natural selection and its’ impacts along the genome, by emphasizing
the intricate mechanisms underlying genetic variation and adaptation, revealing
how natural selection acts on specific genomic regions to shape phenotypic diver-
sity and ecological interactions. I hope the findings in my research inspire further
investigation and innovation in the field of evolutionary biology, fostering a deeper
understanding of the mechanisms driving biological diversity and adaptation. The
threads left behind by evolution are complex and messy, but they are ready for re-
searchers to grasp.
77
APPENDIX A
COEVOLUTION APPENDIX
78
Figure A.1. Newt and snake population sizes for the sixteen genetic architecture
combinations of Experiment 1 over generations 5000 to 10000. Points are colored
by the difference between average snake and average newt phenotypes. In each plot
the population sizes at a evenly spaced set of time points are shown for each of
the four replicates (replicate ID shown by point type). Newt and snake population
sizes are generally negatively correlated: when newts have higher phenotypes (red
points), newt populations tend to be larger and snake populations smaller. Note
there is substantial variation between replicates, probably due to differing initial
conditions.
79
Figure A.2. As in Figure A.1, except that time points are between 45,000 and
50,000 generations (the end of the simulations). Note that there is less variation
between replicates than in the earlier time points.
80
Figure A.3. Distributions of spatial correlations between newt and snake phe-
notypes, across the combinations of genetic architectures of Experiment 1. Each
boxplot shows the range of spatial correlations computed across an evenly spaced
set of time points from 40,000 to 50,000 generations. (top) “Heterogeneous cost -
no interaction” refers to simulations where costliness for both species varies across
the landscape, but interaction outcome does not depend on phenotype (i.e., is just
the result of a coin toss); (middle) “Heterogeneous cost - snake” refers to simu-
lations that are as usual except that the costliness of the snake phenotype varies
across the landscape (but not newt phenotypes); (bottom) “Heterogeneous cost -
newt” refers to the converse situation, where newt phenotype costliness varies, but
not snake.
81
Figure A.4. Distribution of mean newt and snake phenotypes and mean (snake
minus newt) phenotype differences from a regularly spaced set of generations be-
tween 45,000 and 50,000. in Experiment 2, except for those simulations including
a genetic architecture with the lowest mutational variance (1b, 1c, 1d, or 1e). (A)
shows all nine combinations of mutational variances in the first shaded section of
Experiment 2 of Table 2.1 (i.e., genetic architectures 2b, 3b, and 4b). The com-
bination is shown on the x-axis labels, prepended with the difference of snake
and newt log10(VM) values: for instance, the leftmost set of boxplots, labeled “(-
2)S:2b/N:4b”, refers to simulations in which snakes have genetic architecture 2b,
newts have genetic architecture 4b; and the snake’s genetic architecture has 100
times less mutational variance than does the newt’s. (B, C, and D) show the same,
but for combinations in the remaining three sections of Experiment 2 in Table 2.1.
The arrangement is so that for boxplots on the left, snakes have lower mutational
variance than newts, and on the right, newts have lower mutational variance than
snakes.
82
Figure A.5. Mean newt and snake phenotypes (and mean differences), much as in
Figure A.4, but for Experiment 3. In this experiment, the different genetic ar-
chitectures are grouped in Table 2.1 by mutation rate, so the boxplots are ordered
by difference in log10(µ), so that on the left, snakes have lower mutation rate than
newts, and on the right, newts have lower mutation rate than snakes.
83
APPENDIX B
CICHLID APPENDIX
84
85
Table B.1. Samples used in this study (from Malinsky, Svardal, et al. (2018))
Sample MeanCov Accession num Accession num2 Accession num3
Champsochromis caeruelus 14.1176 ERR715493
Champsochromis caeruelus 14.0589 ERR715494
Chilotilapia rhoadesii 13.2928 ERR715491
Chilotilapia rhoadesii 4.96914 ERR702296
Chilotilapia rhoadesii 4.65948 ERR702297
Chilotilapia rhoadesii 5.23728 ERR702308
Copadichromis virginalis 13.4079 ERR299213 ERR299204 ERR299204
Copadichromis virginalis 5.83318 ERR702292
Copadichromis virginalis 5.54251 ERR702293
Copadichromis virginalis 5.84242 ERR702294
Copadichromis virginalis 5.45189 ERR702295
Copadichromis virginalis 4.3907 ERR715482
Copadichromis virginalis 4.40469 ERR715483
Ctenopharynx intermedius 4.69935 ERR702298
Ctenopharynx intermedius 13.1535 ERR715492
Ctenopharynx intermedius 4.64397 ERR702299
Dimidiochromis strigatus 13.1328 ERR1081380
Dimidiochromis strigatus 12.9357 ERR715523
Continues on the next page ...
86
Table B.1. The table (continued from previous page)
Sample MeanCov Accession num Accession num2 Accession num3
Protomelas ornatus 14.4113 ERR715496
Protomelas ornatus 13.7368 ERR715497
Fossorochromis rostratus 4.54727 ERR715476
Fossorochromis rostratus 5.20952 ERR702309
Fossorochromis rostratus 12.5285 ERR715512
Fossorochromis rostratus 13.0272 ERR715516
Hemitilapia oxyrhynchus 12.9473 ERR1081376
Hemitilapia oxyrhynchus 12.0616 ERR715518
Lethrinops lethrinus 11.5057 ERR271663 ERR271655 ERR271655
Lethrinops lethrinus 12.6041 ERR315230 ERR303405 ERR303405
Mylochromis anaphyrmus 12.5896 ERR266461 ERR266493 ERR266493
Mylochromis anaphyrmus 5.41922 ERR702288
Mylochromis anaphyrmus 5.42683 ERR702289
Mylochromis anaphyrmus 5.46714 ERR702290
Mylochromis anaphyrmus 5.24887 ERR702291
Otopharynx speciosus 13.8435 ERR715488
Otopharynx speciosus 12.8562 ERR715489
Placidochromis subocularis 4.92934 ERR702304
Continues on the next page ...
87
Table B.1. The table (continued from previous page)
Sample MeanCov Accession num Accession num2 Accession num3
Placidochromis subocularis 4.42822 ERR715487
Placidochromis subocularis 4.66737 ERR715486
Placidochromis subocularis 4.34062 ERR715485
Placidochromis subocularis 4.34459 ERR715484
Placidochromis subocularis 5.1356 ERR702307
Placidochromis subocularis 5.12527 ERR702306
Placidochromis cf longimanus 12.6469 ERR715524
Placidochromis cf longimanus 4.63765 ERR715478
Placidochromis cf longimanus 4.54243 ERR715479
Placidochromis cf longimanus 4.50002 ERR715480
Placidochromis cf longimanus 4.49001 ERR715481
Protomelas ornatus 14.0089 ERR715514
Protomelas ornatus 13.0325 ERR1081372
Tremitochranus placodon 4.91619 ERR702300
Tremitochranus placodon 4.97781 ERR702301
Tremitochranus placodon 4.82966 ERR702302
Tremitochranus placodon 5.18536 ERR702303
C.virginalis4
C.virginalis2
C.virginalis6
C.virginalis7
C.virginalis5
C.virginalis3
C.virginalis1
L.lethrinus3*
L.lethrinus1
D.strigatus2
D.strigatus1
P.subocularis6
P.subocularis1
O.speciosus2
O.speciosus1
H.oxyrhynchus2
H.oxyrhynchus1
P.ornatus4*
P.ornatus3*
P.ornatus2
P.ornatus1
C caeruelus2
C.caeruelus1
C.intermedius2
C.intermedius3
C.intermedius1
F.rostratus3
F.rostratus1
F.rostratus4
F.rostratus2
T.placodon4
T.placodon2
T.placodon3
T.placodon1
C.rhoadesii4
C.rhoadesii3
C.rhoadesii2
C.rhoadesii1
P.longimanus3
P.longimanus1
P.longimanus4
P.longimanus5
P.longimanus2
P.subocularis7
P.subocularis4
P.subocularis5
P.subocularis3
P.subocularis2
M.anaphyrmus5
M.anaphyrmus4
M.anaphyrmus2
M.anaphyrmus3
M.anaphyrmus1
Figure B.1. Phylogenetic tree of all individuals (see text for details).
88
A Pi by Window
0.006
0.005
0.004
0.003
0.002
0.001
0.000
0 200 400 600 800
Window
B LG 5: Pi By Window C LG 11: Pi By Window D LG 22: Pi By Window
0.003 0.006
0.0020 0.005
0.002 0.004
0.0015
0.003
0.001 0.0010
0.002
0.001
0.0005
150 160 170 370 380 390 700 710 720
Window Window Window
Figure B.2. (A) Genetic diversity along the genome in 1Mb windows. (B, C, D)
Diversity in 1Mb windows for linkage groups 5, 11, and 22.
89
Pi Pi
Pi
Pi
M.anaphyrmus
P.longimanus
L.lethrinus
O.speciosus
C.intermedius
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure B.3. Whole-genome phylogenetic tree for one individual per species (see text
for details).
90
Correlations for each Fish Pair
0.5
Line
Gene Den by Acc
Gene Rep by Acc
0.0
Gene Rep by Gene Den
Rec by Acc
Rec by Gene Den
Rec by Gene Rep
Correlation To Dxy
Accesibility
Gene Density
Gene Repeats
−0.5
Recombination
−1.0
 
Figure B.4. Whole genome correlations for all species pairs. Each dot is a corre-
lation between dXY of a species pair and accessibility (blue), gene density (green),
gene repeat content (orange), or recombination rate (pink). Each line is a cor-
relation between genome functions (gene density by accessibility is a light green
solid line, gene repeat by accessibility is a dashed purple line, gene repeat by gene
density is an orange dotted line, recombination rate by accessibility is light green
dashed and dotted line, recombination rate by gene density is red long dashed line,
and recombination rate by gene repeat is a blue two dashed line).
91
Correlation
C.caeruelus C.virginalis
C.caeruelus D.strigatus
C.caeruelus F.rostratus
C.caeruelus H.oxyrhynchus
C.caeruelus L.lethrinus
C.caeruelus O.speciosus
C.caeruelus P.ornatus
C.intermedius C.caeruelus
C.intermedius C.rhoadesii
C.intermedius C.virginalis
C.intermedius D.strigatus
C.intermedius F.rostratus
C.intermedius H.oxyrhynchus
C.intermedius L.lethrinus
C.intermedius O.speciosus
C.intermedius P.ornatus
C.intermedius P.subocularis
C.rhoadesii C.caeruelus
C.rhoadesii C.virginalis
C.rhoadesii D.strigatus
C.rhoadesii F.rostratus
C.rhoadesii H.oxyrhynchus
C.rhoadesii L.lethrinus
C.rhoadesii O.speciosus
C.rhoadesii P.ornatus
C.virginalis D.strigatus
C.virginalis F.rostratus
C.virginalis H.oxyrhynchus
C.virginalis L.lethrinus
C.virginalis O.speciosus
C.virginalis P.ornatus
D.strigatus F.rostratus
D.strigatus H.oxyrhynchus
D.strigatus L.lethrinus
D.strigatus O.speciosus
D.strigatus P.ornatus
F.rostratus H.oxyrhynchus
F.rostratus L.lethrinus
F.rostratus O.speciosus
H.oxyrhynchus L.lethrinus
H.oxyrhynchus O.speciosus
L.lethrinus O.speciosus
M.anaphyrmus C.caeruelus
M.anaphyrmus C.intermedius
M.anaphyrmus C.rhoadesii
M.anaphyrmus C.virginalis
M.anaphyrmus D.strigatus
M.anaphyrmus F.rostratus
M.anaphyrmus H.oxyrhynchus
M.anaphyrmus L.lethrinus
M.anaphyrmus O.speciosus
M.anaphyrmus P.longimanus
M.anaphyrmus P.ornatus
M.anaphyrmus P.subocularis
M.anaphyrmus T.placodon
P.longimanus C.caeruelus
P.longimanus C.intermedius
P.longimanus C.rhoadesii
P.longimanus C.virginalis
P.longimanus D.strigatus
P.longimanus F.rostratus
P.longimanus H.oxyrhynchus
P.longimanus L.lethrinus
P.longimanus O.speciosus
P.longimanus P.ornatus
P.longimanus P.subocularis
P.ornatus F.rostratus
P.ornatus H.oxyrhynchus
P.ornatus L.lethrinus
P.ornatus O.speciosus
P.subocularis C.caeruelus
P.subocularis C.rhoadesii
P.subocularis C.virginalis
P.subocularis D.strigatus
P.subocularis F.rostratus
P.subocularis H.oxyrhynchus
P.subocularis L.lethrinus
P.subocularis O.speciosus
P.subocularis P.ornatus
T.placodon C.caeruelus
T.placodon C.intermedius
T.placodon C.rhoadesii
T.placodon C.virginalis
T.placodon D.strigatus
T.placodon F.rostratus
T.placodon H.oxyrhynchus
T.placodon L.lethrinus
T.placodon O.speciosus
T.placodon P.longimanus
T.placodon P.ornatus
T.placodon P.subocularis
great apes
exon
recrate
5 10 15 20
chromosome
Figure B.5. Correlations of landscapes of diversity and divergence in the great apes
with recombination rate (“recrate”, red) and exon density (“exon”), by chromo-
some. Data are re-plotted from Rodrigues, Kern, and Ralph (2024): correlations
are Spearman correlations for 1Mb windows, and are shown for genetic diversity
and divergence within and between humans, bonobo, chimpanzee, gorilla, and
orangutan.
92
Spearman correlation
−0.5 0.0 0.5
Figure B.6. Each point is a 1Mb window of admixture proportions (fd, trios in 3.5).
(A) Whole genome fd values by dXY between rostratus and caeruelus. (B) Whole
genome fd values by dXY between caeruelus and speciosus. (C) Whole genome fd
values by π speciosus. (D) Whole genome fd values by π caeruelus.
93
APPENDIX C
WINDOW DXY BY SPECIES
94
C.intermedius LG 1 C.intermedius LG 2 C.intermedius LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.intermedius LG 4 C.intermedius LG 5 C.intermedius LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.intermedius LG 7 C.intermedius LG 8 C.intermedius LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.intermedius LG 10 C.intermedius LG 11 C.intermedius LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.intermedius LG 13 C.intermedius LG 14 C.intermedius LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.intermedius LG 16 C.intermedius LG 17 C.intermedius LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.intermedius LG 19 C.intermedius LG 20 C.intermedius LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.intermedius LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.1. 1Mb window dXY for all linkage groups, C.intermedius
95
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
M.anaphyrmus LG 1 M.anaphyrmus LG 2 M.anaphyrmus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
M.anaphyrmus LG 4 M.anaphyrmus LG 5 M.anaphyrmus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
M.anaphyrmus LG 7 M.anaphyrmus LG 8 M.anaphyrmus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
M.anaphyrmus LG 10 M.anaphyrmus LG 11 M.anaphyrmus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
M.anaphyrmus LG 13 M.anaphyrmus LG 14 M.anaphyrmus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
M.anaphyrmus LG 16 M.anaphyrmus LG 17 M.anaphyrmus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
M.anaphyrmus LG 19 M.anaphyrmus LG 20 M.anaphyrmus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
M.anaphyrmus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.2. 1Mb window dXY for all linkage groups, M.anaphyrmus
96
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
P.subocularis LG 1 P.subocularis LG 2 P.subocularis LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.subocularis LG 4 P.subocularis LG 5 P.subocularis LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.subocularis LG 7 P.subocularis LG 8 P.subocularis LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.subocularis LG 10 P.subocularis LG 11 P.subocularis LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.subocularis LG 13 P.subocularis LG 14 P.subocularis LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.subocularis LG 16 P.subocularis LG 17 P.subocularis LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.subocularis LG 19 P.subocularis LG 20 P.subocularis LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.subocularis LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.3. 1Mb window dXY for all linkage groups, P.subocularis
97
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
C.rhoadesii LG 1 C.rhoadesii LG 2 C.rhoadesii LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.rhoadesii LG 4 C.rhoadesii LG 5 C.rhoadesii LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.rhoadesii LG 7 C.rhoadesii LG 8 C.rhoadesii LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.rhoadesii LG 10 C.rhoadesii LG 11 C.rhoadesii LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.rhoadesii LG 13 C.rhoadesii LG 14 C.rhoadesii LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.rhoadesii LG 16 C.rhoadesii LG 17 C.rhoadesii LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.rhoadesii LG 19 C.rhoadesii LG 20 C.rhoadesii LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.rhoadesii LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.4. 1Mb window dXY for all linkage groups, C.rhoadesii
98
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
C.caeruelus LG 1 C.caeruelus LG 2 C.caeruelus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.caeruelus LG 4 C.caeruelus LG 5 C.caeruelus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.caeruelus LG 7 C.caeruelus LG 8 C.caeruelus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.caeruelus LG 10 C.caeruelus LG 11 C.caeruelus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.caeruelus LG 13 C.caeruelus LG 14 C.caeruelus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.caeruelus LG 16 C.caeruelus LG 17 C.caeruelus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.caeruelus LG 19 C.caeruelus LG 20 C.caeruelus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.caeruelus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.5. 1Mb window dXY for all linkage groups, C.caeruelus
99
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
D.strigatus LG 1 D.strigatus LG 2 D.strigatus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
D.strigatus LG 4 D.strigatus LG 5 D.strigatus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
D.strigatus LG 7 D.strigatus LG 8 D.strigatus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
D.strigatus LG 10 D.strigatus LG 11 D.strigatus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
D.strigatus LG 13 D.strigatus LG 14 D.strigatus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
D.strigatus LG 16 D.strigatus LG 17 D.strigatus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
D.strigatus LG 19 D.strigatus LG 20 D.strigatus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
D.strigatus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.6. 1Mb window dXY for all linkage groups, D.strigatus
100
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
P.ornatus LG 1 P.ornatus LG 2 P.ornatus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.ornatus LG 4 P.ornatus LG 5 P.ornatus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.ornatus LG 7 P.ornatus LG 8 P.ornatus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.ornatus LG 10 P.ornatus LG 11 P.ornatus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.ornatus LG 13 P.ornatus LG 14 P.ornatus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.ornatus LG 16 P.ornatus LG 17 P.ornatus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.ornatus LG 19 P.ornatus LG 20 P.ornatus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.ornatus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.7. 1Mb window dXY for all linkage groups, P.ornatus
101
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
F.rostratus LG 1 F.rostratus LG 2 F.rostratus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
F.rostratus LG 4 F.rostratus LG 5 F.rostratus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
F.rostratus LG 7 F.rostratus LG 8 F.rostratus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
F.rostratus LG 10 F.rostratus LG 11 F.rostratus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
F.rostratus LG 13 F.rostratus LG 14 F.rostratus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
F.rostratus LG 16 F.rostratus LG 17 F.rostratus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
F.rostratus LG 19 F.rostratus LG 20 F.rostratus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
F.rostratus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.8. 1Mb window dXY for all linkage groups, F.rostratus
102
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
H.oxyrhynchus LG 1 H.oxyrhynchus LG 2 H.oxyrhynchus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
H.oxyrhynchus LG 4 H.oxyrhynchus LG 5 H.oxyrhynchus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
H.oxyrhynchus LG 7 H.oxyrhynchus LG 8 H.oxyrhynchus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
H.oxyrhynchus LG 10 H.oxyrhynchus LG 11 H.oxyrhynchus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
H.oxyrhynchus LG 13 H.oxyrhynchus LG 14 H.oxyrhynchus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
H.oxyrhynchus LG 16 H.oxyrhynchus LG 17 H.oxyrhynchus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
H.oxyrhynchus LG 19 H.oxyrhynchus LG 20 H.oxyrhynchus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
H.oxyrhynchus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.9. 1Mb window dXY for all linkage groups, H.oxyrhynchus
103
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
L.lethrinus LG 1 L.lethrinus LG 2 L.lethrinus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
L.lethrinus LG 4 L.lethrinus LG 5 L.lethrinus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
L.lethrinus LG 7 L.lethrinus LG 8 L.lethrinus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
L.lethrinus LG 10 L.lethrinus LG 11 L.lethrinus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
L.lethrinus LG 13 L.lethrinus LG 14 L.lethrinus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
L.lethrinus LG 16 L.lethrinus LG 17 L.lethrinus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
L.lethrinus LG 19 L.lethrinus LG 20 L.lethrinus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
L.lethrinus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.10. 1Mb window dXY for all linkage groups, L.lethrinus
104
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
C.virginalis LG 1 C.virginalis LG 2 C.virginalis LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.virginalis LG 4 C.virginalis LG 5 C.virginalis LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.virginalis LG 7 C.virginalis LG 8 C.virginalis LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.virginalis LG 10 C.virginalis LG 11 C.virginalis LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.virginalis LG 13 C.virginalis LG 14 C.virginalis LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.virginalis LG 16 C.virginalis LG 17 C.virginalis LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.virginalis LG 19 C.virginalis LG 20 C.virginalis LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.virginalis LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.11. 1Mb window dXY for all linkage groups, C.virginalis
105
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
O.speciosus LG 1 O.speciosus LG 2 O.speciosus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
O.speciosus LG 4 O.speciosus LG 5 O.speciosus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
O.speciosus LG 7 O.speciosus LG 8 O.speciosus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
O.speciosus LG 10 O.speciosus LG 11 O.speciosus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
O.speciosus LG 13 O.speciosus LG 14 O.speciosus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
O.speciosus LG 16 O.speciosus LG 17 O.speciosus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
O.speciosus LG 19 O.speciosus LG 20 O.speciosus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
O.speciosus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.12. 1Mb window dXY for all linkage groups, O.speciosus
106
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
T.placodon LG 1 T.placodon LG 2 T.placodon LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
T.placodon LG 4 T.placodon LG 5 T.placodon LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
T.placodon LG 7 T.placodon LG 8 T.placodon LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
T.placodon LG 10 T.placodon LG 11 T.placodon LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
T.placodon LG 13 T.placodon LG 14 T.placodon LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
T.placodon LG 16 T.placodon LG 17 T.placodon LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
T.placodon LG 19 T.placodon LG 20 T.placodon LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
T.placodon LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.13. 1Mb window dXY for all linkage groups, T.placodon
107
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
P.longimanus LG 1 P.longimanus LG 2 P.longimanus LG 3
0.004
0.0020 0.004
0.003
0.0016 0.003
0.002
0.0012
0.001 0.002
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.longimanus LG 4 P.longimanus LG 5 P.longimanus LG 6
0.005 0.0030
0.0030
0.004 0.0025 0.0025
0.003 0.0020 0.0020
0.002 0.0015 0.0015
0.001 0.0010
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.longimanus LG 7 P.longimanus LG 8 P.longimanus LG 9
0.0030
0.0025
0.0025 0.003
0.0020
0.0020
0.0015 0.002
0.0010 0.0015
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.longimanus LG 10 P.longimanus LG 11 P.longimanus LG 12
0.0025 0.0030 0.0025
0.0025
0.0020 0.0020
0.0020
0.0015 0.00150.0015
0.0010 0.0010 0.0010
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.longimanus LG 13 P.longimanus LG 14 P.longimanus LG 15
0.0030 0.0025
0.0025 0.003
0.0020
0.0020 0.002
0.0015 0.0015
0.0010 0.001
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.longimanus LG 16 P.longimanus LG 17 P.longimanus LG 18
0.0030
0.0025 0.0021
0.0025
0.0020 0.0018
0.0020
0.0015 0.0015
0.0015
0.0010
0.0012 0.0010
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.longimanus LG 19 P.longimanus LG 20 P.longimanus LG 22
0.0030 0.006
0.0025
0.0025
0.0020 0.0040.0020
0.0015 0.0015 0.002
0.0010 0.0010
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.longimanus LG 23
0.003
0.002
0.001
730 740 750 760 770
Window
Figure C.14. 1Mb window dXY for all linkage groups, P.longimanus
108
Dxy Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
Dxy Dxy Dxy Dxy Dxy Dxy Dxy
APPENDIX D
WINDOW π BY SPECIES
109
P.longimanus LG 1 P.longimanus LG 2 P.longimanus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.longimanus LG 4 P.longimanus LG 5 P.longimanus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.longimanus LG 7 P.longimanus LG 8 P.longimanus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.longimanus LG 10 P.longimanus LG 11 P.longimanus LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.longimanus LG 13 P.longimanus LG 14 P.longimanus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.longimanus LG 16 P.longimanus LG 17 P.longimanus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.longimanus LG 19 P.longimanus LG 20 P.longimanus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.longimanus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.1. 1Mb window π for all linkage groups, C.intermedius
110
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
C.intermedius LG 1 C.intermedius LG 2 C.intermedius LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.intermedius LG 4 C.intermedius LG 5 C.intermedius LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.intermedius LG 7 C.intermedius LG 8 C.intermedius LG 9
0.0025
0.0030
0.0020 0.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.intermedius LG 10 C.intermedius LG 11 C.intermedius LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.intermedius LG 13 C.intermedius LG 14 C.intermedius LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.intermedius LG 16 C.intermedius LG 17 C.intermedius LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.intermedius LG 19 C.intermedius LG 20 C.intermedius LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.intermedius LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.2. 1Mb window π for all linkage groups, M.anaphyrmus
111
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
T.placodon LG 1 T.placodon LG 2 T.placodon LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
T.placodon LG 4 T.placodon LG 5 T.placodon LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
T.placodon LG 7 T.placodon LG 8 T.placodon LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
T.placodon LG 10 T.placodon LG 11 T.placodon LG 12
0.0025
0.0020 0.0020
0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
T.placodon LG 13 T.placodon LG 14 T.placodon LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
T.placodon LG 16 T.placodon LG 17 T.placodon LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
T.placodon LG 19 T.placodon LG 20 T.placodon LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
T.placodon LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.3. 1Mb window π for all linkage groups, P.subocularis
112
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
P.subocularis LG 1 P.subocularis LG 2 P.subocularis LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.subocularis LG 4 P.subocularis LG 5 P.subocularis LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.subocularis LG 7 P.subocularis LG 8 P.subocularis LG 9
0.0025
0.0030
0.0020 0.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.subocularis LG 10 P.subocularis LG 11 P.subocularis LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.subocularis LG 13 P.subocularis LG 14 P.subocularis LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.subocularis LG 16 P.subocularis LG 17 P.subocularis LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.subocularis LG 19 P.subocularis LG 20 P.subocularis LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.subocularis LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.4. 1Mb window π for all linkage groups, C.rhoadesii
113
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
C.virginalis LG 1 C.virginalis LG 2 C.virginalis LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.virginalis LG 4 C.virginalis LG 5 C.virginalis LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.virginalis LG 7 C.virginalis LG 8 C.virginalis LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.virginalis LG 10 C.virginalis LG 11 C.virginalis LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.virginalis LG 13 C.virginalis LG 14 C.virginalis LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.virginalis LG 16 C.virginalis LG 17 C.virginalis LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.virginalis LG 19 C.virginalis LG 20 C.virginalis LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.virginalis LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.5. 1Mb window π for all linkage groups, C.caeruelus
114
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
C.caeruelus LG 1 C.caeruelus LG 2 C.caeruelus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.caeruelus LG 4 C.caeruelus LG 5 C.caeruelus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.caeruelus LG 7 C.caeruelus LG 8 C.caeruelus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.caeruelus LG 10 C.caeruelus LG 11 C.caeruelus LG 12
0.0025
0.0020 0.0020
0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.caeruelus LG 13 C.caeruelus LG 14 C.caeruelus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.caeruelus LG 16 C.caeruelus LG 17 C.caeruelus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.caeruelus LG 19 C.caeruelus LG 20 C.caeruelus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.caeruelus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.6. 1Mb window π for all linkage groups, D.strigatus
115
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
P.ornatus LG 1 P.ornatus LG 2 P.ornatus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
P.ornatus LG 4 P.ornatus LG 5 P.ornatus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
P.ornatus LG 7 P.ornatus LG 8 P.ornatus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
P.ornatus LG 10 P.ornatus LG 11 P.ornatus LG 12
0.0025
0.0020 0.00200.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
P.ornatus LG 13 P.ornatus LG 14 P.ornatus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
P.ornatus LG 16 P.ornatus LG 17 P.ornatus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
P.ornatus LG 19 P.ornatus LG 20 P.ornatus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
P.ornatus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.7. 1Mb window π for all linkage groups, P.ornatus
116
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
D.strigatus LG 1 D.strigatus LG 2 D.strigatus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
D.strigatus LG 4 D.strigatus LG 5 D.strigatus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
D.strigatus LG 7 D.strigatus LG 8 D.strigatus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
D.strigatus LG 10 D.strigatus LG 11 D.strigatus LG 12
0.0025
0.0020 0.0020
0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
D.strigatus LG 13 D.strigatus LG 14 D.strigatus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
D.strigatus LG 16 D.strigatus LG 17 D.strigatus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
D.strigatus LG 19 D.strigatus LG 20 D.strigatus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
D.strigatus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.8. 1Mb window π for all linkage groups, F.rostratus
117
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
F.rostratus LG 1 F.rostratus LG 2 F.rostratus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
F.rostratus LG 4 F.rostratus LG 5 F.rostratus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
F.rostratus LG 7 F.rostratus LG 8 F.rostratus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
F.rostratus LG 10 F.rostratus LG 11 F.rostratus LG 12
0.0025
0.0020 0.00200.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
F.rostratus LG 13 F.rostratus LG 14 F.rostratus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
F.rostratus LG 16 F.rostratus LG 17 F.rostratus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
F.rostratus LG 19 F.rostratus LG 20 F.rostratus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
F.rostratus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.9. 1Mb window π for all linkage groups, H.oxyrhynchus
118
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
H.oxyrhynchus LG 1 H.oxyrhynchus LG 2 H.oxyrhynchus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
H.oxyrhynchus LG 4 H.oxyrhynchus LG 5 H.oxyrhynchus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
H.oxyrhynchus LG 7 H.oxyrhynchus LG 8 H.oxyrhynchus LG 9
0.0025
0.0030
0.0020 0.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
H.oxyrhynchus LG 10 H.oxyrhynchus LG 11 H.oxyrhynchus LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
H.oxyrhynchus LG 13 H.oxyrhynchus LG 14 H.oxyrhynchus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
H.oxyrhynchus LG 16 H.oxyrhynchus LG 17 H.oxyrhynchus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
H.oxyrhynchus LG 19 H.oxyrhynchus LG 20 H.oxyrhynchus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
H.oxyrhynchus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.10. 1Mb window π for all linkage groups, L.lethrinus
119
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
L.lethrinus LG 1 L.lethrinus LG 2 L.lethrinus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
L.lethrinus LG 4 L.lethrinus LG 5 L.lethrinus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
L.lethrinus LG 7 L.lethrinus LG 8 L.lethrinus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
L.lethrinus LG 10 L.lethrinus LG 11 L.lethrinus LG 12
0.0025
0.0020 0.00200.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
L.lethrinus LG 13 L.lethrinus LG 14 L.lethrinus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
L.lethrinus LG 16 L.lethrinus LG 17 L.lethrinus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
L.lethrinus LG 19 L.lethrinus LG 20 L.lethrinus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
L.lethrinus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.11. 1Mb window π for all linkage groups, C.virginalis
120
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
O.speciosus LG 1 O.speciosus LG 2 O.speciosus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
O.speciosus LG 4 O.speciosus LG 5 O.speciosus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
O.speciosus LG 7 O.speciosus LG 8 O.speciosus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
O.speciosus LG 10 O.speciosus LG 11 O.speciosus LG 12
0.0025
0.0020 0.00200.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
O.speciosus LG 13 O.speciosus LG 14 O.speciosus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
O.speciosus LG 16 O.speciosus LG 17 O.speciosus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
O.speciosus LG 19 O.speciosus LG 20 O.speciosus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
O.speciosus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.12. 1Mb window π for all linkage groups, O.speciosus
121
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
C.rhoadesii LG 1 C.rhoadesii LG 2 C.rhoadesii LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
C.rhoadesii LG 4 C.rhoadesii LG 5 C.rhoadesii LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
C.rhoadesii LG 7 C.rhoadesii LG 8 C.rhoadesii LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
C.rhoadesii LG 10 C.rhoadesii LG 11 C.rhoadesii LG 12
0.0025
0.0020 0.0020
0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
C.rhoadesii LG 13 C.rhoadesii LG 14 C.rhoadesii LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
C.rhoadesii LG 16 C.rhoadesii LG 17 C.rhoadesii LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
C.rhoadesii LG 19 C.rhoadesii LG 20 C.rhoadesii LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
C.rhoadesii LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.13. 1Mb window π for all linkage groups, T.placodon
122
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
M.anaphyrmus LG 1 M.anaphyrmus LG 2 M.anaphyrmus LG 3
0.0020 0.004
0.003
0.0015 0.003
0.002
0.0010 0.002
0.0005 0.001 0.001
0 10 20 30 40 40 50 60 70 80 90 100 110
Window Window Window
M.anaphyrmus LG 4 M.anaphyrmus LG 5 M.anaphyrmus LG 6
0.003 0.003
0.004
0.003 0.002 0.002
0.002
0.001 0.001
0.001
110 120 130 140 150 160 170 180 190 200 210
Window Window Window
M.anaphyrmus LG 7 M.anaphyrmus LG 8 M.anaphyrmus LG 9
0.0025
0.0020 0.00300.0020
0.0025
0.0015 0.0015 0.0020
0.0010 0.0010 0.0015
0.0005 0.0010
0.0005
220 240 260 280 290 300 310 315 320 325
Window Window Window
M.anaphyrmus LG 10 M.anaphyrmus LG 11 M.anaphyrmus LG 12
0.0025
0.0020
0.0020 0.0020
0.0015 0.0015 0.0015
0.0010 0.0010 0.0010
0.0005 0.0005
0.0005
330 340 350 360 370 380 390 400 410 420 430
Window Window Window
M.anaphyrmus LG 13 M.anaphyrmus LG 14 M.anaphyrmus LG 15
0.0020 0.0020 0.003
0.0015 0.0015 0.002
0.0010 0.0010
0.001
0.0005 0.0005
430 440 450 460 470 480 490 500 510 520 530
Window Window Window
M.anaphyrmus LG 16 M.anaphyrmus LG 17 M.anaphyrmus LG 18
0.0025 0.0020
0.0020
0.0015 0.002
0.0015
0.0010 0.0010
0.001
0.0005 0.0005
540 550 560 570 580 590 600 610 620 630
Window Window Window
M.anaphyrmus LG 19 M.anaphyrmus LG 20 M.anaphyrmus LG 22
0.006
0.0025
0.005
0.0020 0.002 0.004
0.0015 0.003
0.0010 0.001 0.002
0.001
0.0005
640 645 650 655 660 670 680 690 700 710 720
Window Window Window
M.anaphyrmus LG 23
0.002
0.001
730 740 750 760 770
Window
Figure D.14. 1Mb window π for all linkage groups, P.longimanus
123
pi pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
pi pi pi pi pi pi pi
APPENDIX E
CM BY BASE PAIR
LG 1 updated file
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.1. cM by base pair linkage group one
124
cM
LG 2 updated file
90
80
70
60
50
40
0e+00 1e+07 2e+07 3e+07
BP
Figure E.2. cM by base pair linkage group two
LG 3 updated file
100
80
60
40
1e+07 2e+07 3e+07
BP
Figure E.3. cM by base pair linkage group three
125
cM cM
LG 4 updated file
50
40
30
20
10
0
1e+07 2e+07 3e+07
BP
Figure E.4. cM by base pair linkage group four
LG 5 updated file
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.5. cM by base pair linkage group five
126
cM cM
LG 6 updated file
40
30
20
10
0
0e+00 1e+07 2e+07 3e+07 4e+07
BP
Figure E.6. cM by base pair linkage group six
LG 7 updated file
100
75
50
25
0
2e+07 4e+07 6e+07
BP
Figure E.7. cM by base pair linkage group seven
127
cM cM
LG 8 updated file
50
40
30
20
10
0
0.0e+00 5.0e+06 1.0e+07 1.5e+07 2.0e+07
BP
Figure E.8. cM by base pair linkage group eight
LG 9 updated file
40
30
20
10
0
5.0e+06 1.0e+07 1.5e+07 2.0e+07
BP
Figure E.9. cM by base pair linkage group nine
128
cM cM
LG 10 updated file
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.10. cM by base pair linkage group ten
LG 11 updated file
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.11. cM by base pair linkage group eleven
129
cM cM
LG 12 updated file
75
50
25
0
1e+07 2e+07 3e+07
BP
Figure E.12. cM by base pair linkage group twelve
LG 13 updated file
80
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.13. cM by base pair linkage group thirteen
130
cM cM
LG 14 updated file
75
50
25
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.14. cM by base pair linkage group fourteen
LG 15 updated file
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.15. cM by base pair linkage group fifteen
131
cM cM
LG 16 updated file
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.16. cM by base pair linkage group sixteen
LG 17 updated file
80
60
40
20
0
1e+07 2e+07 3e+07
BP
Figure E.17. cM by base pair linkage group seventeen
132
cM cM
LG 18 updated file
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.18. cM by base pair linkage group eighteen
LG 19 updated file
80
60
40
20
5.0e+06 1.0e+07 1.5e+07 2.0e+07 2.5e+07
BP
Figure E.19. cM by base pair linkage group nineteen
133
cM cM
LG 20 updated file
80
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.20. cM by base pair linkage group twenty
LG 22 updated file
80
60
40
20
0
0e+00 1e+07 2e+07 3e+07
BP
Figure E.21. cM by base pair linkage group twenty-two
134
cM cM
LG 23 updated file
75
50
25
0
0e+00 1e+07 2e+07 3e+07 4e+07
BP
Figure E.22. cM by base pair linkage group twenty-three
135
cM
APPENDIX F
CORRELATION AND COVARIANCE BY LINKAGE GROUPS
136
Simpletree Cor: Pi vs Pi
1 2 3 4 5
1.00
0.75
0.50
0.25
6 7 8 9 10
1.00
0.75
0.50
0.25 factor(LG)
1 12
11 12 13 14 15 2 13
1.00
3 14
4 15
0.75
5 16
0.50 6 17
7 18
0.25 8 19
9 20
16 17 18 19 20
10 22
1.00
11 23
0.75
0.50
0.25
0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06
22 23
1.00
0.75
0.50
0.25
0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06
distance
Figure F.1. Correlation between π and π in 1Mb widowed by phylogenetic time, all
linkage groups
137
cor
Simpletree Cov: Pi vs Pi
1 2 3 4 5
6e−07
4e−07
2e−07
0e+00
6 7 8 9 10
6e−07
4e−07
2e−07
factor(LG)
0e+00 1 12
11 12 13 14 15 2 13
6e−07 3 14
4 15
4e−07
5 16
6 17
2e−07
7 18
8 19
0e+00
9 20
16 17 18 19 20
10 22
6e−07
11 23
4e−07
2e−07
0e+00
0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06
22 23
6e−07
4e−07
2e−07
0e+00
0.00 0.02 0.04 0.06 0.00 0.02 0.04 0.06
distance
Figure F.2. Covariance between π and π in 1Mb widowed by phylogenetic time, all
linkage groups
138
cov
Simpletree Cor: Dxy vs Pi
1 2 3 4 5
0.8
0.4
0.0
6 7 8 9 10
0.8
0.4
0.0
factor(LG)
1 12
11 12 13 14 15 2 13
3 14
0.8
4 15
5 16
0.4
6 17
0.0 7 18
8 19
9 20
16 17 18 19 20
10 22
0.8 11 23
0.4
0.0
0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04
22 23
0.8
0.4
0.0
0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04
distance
Figure F.3. Correlation between dXY and π in 1Mb widowed by phylogenetic time,
all linkage groups
139
cor
Simpletree Cov: Dxy vs Pi
1 2 3 4 5
6e−07
4e−07
2e−07
0e+00
6 7 8 9 10
6e−07
4e−07
2e−07
factor(LG)
0e+00
1 12
11 12 13 14 15 2 13
6e−07 3 14
4 15
4e−07
5 16
6 17
2e−07
7 18
0e+00 8 19
9 20
16 17 18 19 20
10 22
6e−07
11 23
4e−07
2e−07
0e+00
0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04
22 23
6e−07
4e−07
2e−07
0e+00
0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04
distance
Figure F.4. Covariance between dXY and π in 1Mb widowed by phylogenetic time,
all linkage groups
140
cov
Simpletree Cor: Dxy vs Dxy
1 2 3 4 5
1.0
0.8
0.6
0.4
0.2
6 7 8 9 10
1.0
0.8
0.6
0.4
factor(LG)
0.2
1 12
11 12 13 14 15 2 13
1.0
3 14
0.8 4 15
5 16
0.6
6 17
0.4 7 18
8 19
0.2
9 20
16 17 18 19 20
10 22
1.0
11 23
0.8
0.6
0.4
0.2
0.004 0.008 0.012 0.016 0.004 0.008 0.012 0.016 0.004 0.008 0.012 0.016
22 23
1.0
0.8
0.6
0.4
0.2
0.004 0.008 0.012 0.016 0.004 0.008 0.012 0.016
distance
Figure F.5. Correlation between dXY and dXY in 1Mb widowed by phylogenetic
time, all linkage groups
141
cor
Simpletree Cov: Dxy vs Dxy
1 2 3 4 5
6e−07
4e−07
2e−07
0e+00
6 7 8 9 10
6e−07
4e−07
2e−07
factor(LG)
0e+00 1 12
11 12 13 14 15 2 13
3 14
6e−07
4 15
4e−07 5 16
6 17
2e−07 7 18
8 19
0e+00
9 20
16 17 18 19 20
10 22
6e−07 11 23
4e−07
2e−07
0e+00
−0.03−0.02−0.010.00 0.01 −0.03−0.02−0.010.00 0.01 −0.03−0.02−0.010.00 0.01
22 23
6e−07
4e−07
2e−07
0e+00
−0.03−0.02−0.010.00 0.01 −0.03−0.02−0.010.00 0.01
distance
Figure F.6. Covariance between dXY and dXY in 1Mb widowed by phylogenetic
time, all linkage groups
142
cov
APPENDIX G
CORRELATION OF dXY AND GENOME FUNCTIONS FOR ALL LINKAGE
GROUPS
143
Cor: gene_repeats
1 2 3 4 5
0.6 0.1 0.25 0.450.45
0.40
0.5 0.0 0.40 0.20 0.35
0.35
−0.1 0.15 0.30
0.4 0.30 0.25
−0.2 0.10
0.20
0.3 0.25
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
6 7 8 9 10
0.35 0.60.30 0.0
0.30 0.25
0.5 0.6
−0.1
0.20 0.4
0.5
0.25 0.15 0.3 −0.2
0.10 0.4
0.20 0.2 factor(LG)−0.3
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 1 12
11 12 13 14 15 2 13
0.35
0.6 0.30 3 14
0.30 0.4 0.3 4 15
0.4 0.25 0.2 0.25 0.2 5 16
0.20 0.1 6 17
0.2 0.15 0.0 0.20 0.0 7 18
0.10
0.0 −0.2 0.15 −0.1
8 19
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 9 20
16 17 18 19 20 10 22
0.70
0.25 0.25 11 23
0.65 0.30
0.20 0.5
0.60 0.20 0.25
0.15
0.55 0.15 0.20 0.4
0.10
0.50
0.10 0.15
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
22 23
0.075 0.55
0.050 0.45
0.025 0.35
0.000 0.25
−0.025 0.15
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
distance
Figure G.1. Correlation between dXY and gene repeats in 1Mb widowed by phylo-
genetic time, all linkage groups
144
cor
Cor: gene_density
1 2 3 4 5
0.1
0.0 0.4 −0.10 0.10
−0.1 0.05 0.1
0.3 −0.15
−0.2 0.00
−0.3 0.2 −0.20 0.0
−0.05
−0.4 −0.25
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
6 7 8 9 10
0.20 0.05
0.1 −0.1
0.15 0.05
−0.2 0.000.0
0.10
−0.1 −0.3 0.00 −0.05
0.05
−0.2 −0.4 −0.05 −0.10
0.00
−0.3 −0.5 −0.10 −0.15 factor(LG)
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 1 12
11 12 13 14 15 2 13
−0.10
−0.1 −0.2 3 140.0 0.0
−0.15 4 15
−0.1 −0.2 −0.3 −0.1 5 16
−0.20
6 17
−0.2 −0.4
−0.25 −0.3 −0.2 7 18
−0.3
−0.30 −0.5 8 19
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 9 20
16 17 18 19 20 10 22
0.1 0.20 0.35 11 230.3
0.30 −0.35
0.0 0.2 0.15
0.25
−0.1 0.1 −0.40
0.0 0.10 0.20
−0.2 −0.45
−0.1 0.15
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
22 23
0.3
0.00 0.2
−0.04 0.1
−0.08 0.0
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
distance
Figure G.2. Correlation between dXY and gene density in 1Mb widowed by phylo-
genetic time, all linkage groups
145
cor
Cor: accesibility_mask_window
1 2 3 4 5
−0.45
−0.35
0.0 −0.55 −0.50
−0.4
−0.40 −0.55
−0.1 −0.60
−0.5 −0.45 −0.60
−0.2 −0.65 −0.65
−0.6 −0.50 −0.70
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
6 7 8 9 10
−0.25 −0.5 −0.5
−0.36 −0.30 0.0
−0.6 −0.6−0.35
−0.40 −0.1
−0.40
−0.7 −0.7−0.44 −0.45 −0.2
−0.50 factor(LG)−0.8
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 1 12
11 12 13 14 15 2 13
−0.25 0.2 −0.30 −0.1−0.2 3 14
−0.30 0.0 4 15−0.33 −0.2
−0.4 −0.35 5 16
−0.2 −0.36 −0.3
−0.40 6 17
−0.6 −0.4 −0.4 7 18
−0.45 −0.39
−0.5
−0.8 −0.6 8 19
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 9 20
16 17 18 19 20 10 22
−0.50 −0.25 −0.50 11 23
−0.55 −0.30−0.30 −0.2 −0.55
−0.60 −0.35 −0.35 −0.60
−0.65 −0.3−0.40 −0.40 −0.65
−0.70
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
22 23
−0.400
−0.425 −0.5
−0.450
−0.6
−0.475
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
distance
Figure G.3. Correlation between dXY and accessibility in 1Mb widowed by phyloge-
netic time, all linkage groups
146
cor
Cor: recombination_window
1 2 3 4 5
0.55 0.1 −0.04 0.00
0.50 −0.05
0.45 0.0 −0.08 −0.05
−0.10
0.40
−0.1 −0.12 −0.10
0.35 −0.15
0.30 −0.2 −0.16 −0.15
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
6 7 8 9 10
0.05 0.05 0.12
0.1
0.00
0.00 −0.3 0.080.0
−0.05
0.04
−0.10 −0.05 −0.4 −0.1
−0.15 −0.2
0.00 factor(LG)
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 1 12
11 12 13 14 15 2 13
−0.14 3 14
0.4 −0.040.3 −0.16 4 15
0.25
−0.18 −0.08 5 16
0.3 0.2
0.00 −0.20
6 17
−0.12
0.2 7 180.1 −0.22
−0.25 8 19−0.16
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 9 20
16 17 18 19 20 10 22
11 23
−0.20 0.4 0.05
0.05 0.25
−0.25 0.3 0.00
0.20
−0.30 0.00 −0.050.2
−0.35 −0.05 −0.10
0.15
0.1
−0.40 −0.15 0.10
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
22 23
0.54
0.52 −0.05
0.50 −0.10
0.48 −0.15
0.46 −0.20
−0.25
0.03 0.04 0.05 0.06 0.07 0.03 0.04 0.05 0.06 0.07
distance
Figure G.4. Correlation between dXY and recombination rate in 1Mb widowed by
phylogenetic time, all linkage groups
147
cor
APPENDIX H
PHYLOGENETIC TREES BY LINKAGE GROUP
148
LG 1
C.intermedius
M.anaphyrmus
P.longimanus
L.lethrinus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
F.rostratus
D.strigatus
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.1. Phylogenetic tree, linkage group 1
149
LG 2
P.longimanus
L.lethrinus
F.rostratus
C.intermedius
M.anaphyrmus
T.placodon
O.speciosus
P.subocularis
P.ornatus
H.oxyrhynchus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.2. Phylogenetic tree, linkage group 2
150
LG 3
M.anaphyrmus
L.lethrinus
C.intermedius
T.placodon
P.longimanus
P.subocularis
H.oxyrhynchus
F.rostratus
P.ornatus
O.speciosus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.05 0.04 0.03 0.02 0.01 0
Figure H.3. Phylogenetic tree, linkage group 3
151
LG 4
C.intermedius
O.speciosus
M.anaphyrmus
T.placodon
P.longimanus
L.lethrinus
F.rostratus
H.oxyrhynchus
P.ornatus
C.rhoadesii
P.subocularis
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.4. Phylogenetic tree, linkage group 4
152
LG 5
C.intermedius
M.anaphyrmus
P.longimanus
L.lethrinus
T.placodon
P.subocularis
H.oxyrhynchus
D.strigatus
P.ornatus
C.rhoadesii
F.rostratus
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.5. Phylogenetic tree, linkage group 5
153
LG 6
M.anaphyrmus
L.lethrinus
C.intermedius
T.placodon
P.longimanus
F.rostratus
D.strigatus
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.6. Phylogenetic tree, linkage group 6
154
LG 7
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
D.strigatus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.03 0.02 0.01 0
Figure H.7. Phylogenetic tree, linkage group 7
155
LG 8
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.8. Phylogenetic tree, linkage group 8
156
LG 9
C.intermedius
M.anaphyrmus
P.longimanus
L.lethrinus
O.speciosus
F.rostratus
D.strigatus
T.placodon
P.subocularis
C.rhoadesii
H.oxyrhynchus
P.ornatus
C.caeruelus
C.virginalis
0.03 0.025 0.02 0.015 0.01 0.005 0
Figure H.9. Phylogenetic tree, linkage group 9
157
LG 10
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
M.anaphyrmus
T.placodon
P.subocularis
D.strigatus
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.10. Phylogenetic tree, linkage group10
158
LG 11
C.intermedius
O.speciosus
M.anaphyrmus
P.longimanus
L.lethrinus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
F.rostratus
D.strigatus
C.caeruelus
C.virginalis
0.07 0.06 0.05 0.04 0.03 0.02 0.01 0
Figure H.11. Phylogenetic tree, linkage group 11
159
LG 12
C.intermedius
P.longimanus
L.lethrinus
O.speciosus
M.anaphyrmus
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.12. Phylogenetic tree, linkage group 12
160
LG 13
P.longimanus
L.lethrinus
F.rostratus
C.intermedius
M.anaphyrmus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.13. Phylogenetic tree, linkage group 13
161
LG 14
M.anaphyrmus
P.longimanus
L.lethrinus
C.intermedius
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.14. Phylogenetic tree, linkage group 14
162
LG 15
M.anaphyrmus
P.longimanus
L.lethrinus
T.placodon
O.speciosus
C.intermedius
F.rostratus
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.15. Phylogenetic tree, linkage group 15
163
LG 16
C.intermedius
M.anaphyrmus
P.longimanus
L.lethrinus
F.rostratus
O.speciosus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.16. Phylogenetic tree, linkage group 16
164
LG 17
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
T.placodon
P.subocularis
D.strigatus
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.17. Phylogenetic tree, linkage group 17
165
LG 17
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
T.placodon
P.subocularis
D.strigatus
H.oxyrhynchus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.18. Phylogenetic tree, linkage group 17
166
LG 18
C.intermedius
M.anaphyrmus
L.lethrinus
F.rostratus
P.longimanus
T.placodon
P.subocularis
H.oxyrhynchus
D.strigatus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.19. Phylogenetic tree, linkage group 18
167
LG 19
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
O.speciosus
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.20. Phylogenetic tree, linkage group 19
168
LG 20
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
T.placodon
P.subocularis
H.oxyrhynchus
D.strigatus
P.ornatus
C.rhoadesii
O.speciosus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.21. Phylogenetic tree, linkage group 20
169
LG 22
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
O.speciosus
F.rostratus
T.placodon
H.oxyrhynchus
P.subocularis
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.22. Phylogenetic tree, linkage group 22
170
LG 23
M.anaphyrmus
C.intermedius
P.longimanus
L.lethrinus
F.rostratus
O.speciosus
T.placodon
P.subocularis
H.oxyrhynchus
P.ornatus
C.rhoadesii
D.strigatus
C.caeruelus
C.virginalis
0.04 0.03 0.02 0.01 0
Figure H.23. Phylogenetic tree, linkage group 23
171
APPENDIX I
FBRANCH
172
Figure I.1. fbranch result whole genome
173
Figure I.2. fbranch result linkage group 1
174
Figure I.3. fbranch result linkage group 2
175
Figure I.4. fbranch result linkage group 3
176
Figure I.5. fbranch result linkage group 4
177
Figure I.6. fbranch result linkage group 5
178
Figure I.7. fbranch result linkage group 6
179
Figure I.8. fbranch result linkage group 7
180
Figure I.9. fbranch result linkage group 8
181
Figure I.10. fbranch result linkage group 9
182
Figure I.11. fbranch result linkage group 10
183
Figure I.12. fbranch result linkage group 11
184
Figure I.13. fbranch result linkage group 12
185
Figure I.14. fbranch result linkage group 13
186
Figure I.15. fbranch result linkage group 14
187
Figure I.16. fbranch result linkage group 15
188
Figure I.17. fbranch result linkage group 16
189
Figure I.18. fbranch result linkage group 17
190
Figure I.19. fbranch result linkage group 18
191
Figure I.20. fbranch result linkage group 19
192
Figure I.21. fbranch result linkage group 20
193
Figure I.22. fbranch result linkage group 22
194
Figure I.23. fbranch result linkage group 23
195